Geriatric Nursing 36 (2015) 111e119

Contents lists available at ScienceDirect

Geriatric Nursing journal homepage: www.gnjournal.com

Feature Article

The effect of a “surveillance nurse” telephone support intervention in a home care program Ronald Kelly, PhD *, Lori Godin, RN, BSN, GNC Fraser Health Authority, BC, Canada

a r t i c l e i n f o

a b s t r a c t

Article history: Received 5 May 2014 Received in revised form 17 November 2014 Accepted 24 November 2014 Available online 26 December 2014

This study is an evaluation of a unique “surveillance nurse” telephone support intervention for community-dwelling elderly individuals in a home care program. A combined propensity-based covariate-matching procedure was used to pair each individual who received the intervention (“treatment” condition, nT ¼ 930) to a similar individual who did not receive the intervention (“control” condition, nC1 ¼ 930) from among a large pool of potential control individuals (nC0 ¼ 4656). The intervention consisted of regularly scheduled telephone calls from a surveillance nurse to proactively assess the individual’s well-being, care plan status, use of and need for services (home support, adult day program, physiotherapy, etc.) and home environment (e.g., informal caregiver support). Treatment and control conditions were compared with respect to four service utilization outcomes: (1) rate of survival in the community before institutionalization in an assisted living or nursing home facility or death, (2) rate of emergency room registrations, (3) rate of acute care hospitalizations, and (4) rate of days in hospital, during home care enrollment. Results indicated a beneficial effect of the surveillance nurse intervention on reducing rate of service utilization by increasing the duration of the home care episode. Crown Copyright Ó 2015 Published by Elsevier Inc. All rights reserved.

Keywords: Telephone nurse Surveillance nurse Telephone support Telephone care RAI-HC Propensity scoring Covariate matching

Introduction and background Health care systems in North America are confronted with three unyielding realities: (1) a shift in age demographics toward an older population, (2) financial pressures that require health care systems to maintain or expand program capacity while holding or even reducing costs, and (3) the phenomenon that older adults are disproportionate users of health care system resources, with greater risks of mortality and higher rates of multimorbidity. In the Canadian Province of British Columbia (BC) for example, the total number of individuals aged 65 years and older is currently estimated at over 750,000.1 As a percentage of total population, this age group has grown steadily in BC over the past 40 years, from 9.3% (1971), to 11.6% (1985), to 13.1% (2000), to its current level of 16.9%. It is projected that by 2036, almost 25% of the population of the province, approximately 1.5 million people, will be 65 years of age or older.2 Meanwhile, total health spending in BC will reach $19.6 billion by the fiscal year 2016/17, more than 42% of all government expenditures.3 Finally, a recent Canadian Institute of Health Information report found that although individuals aged 65 years and

* Corresponding author. # 504 e 3292 Production Way, Burnaby, BC, Canada V5A 4R4. Tel.: þ1 604 415 8700x538743; fax: þ1 604 415 8701. E-mail address: [email protected] (R. Kelly).

older comprised only 14% of the Canadian population in 2009/10, they accounted for 40% of all hospitalizations and had hospital stays 1.5 times longer than non-seniors.4 Clearly, there is an imperative to provide effective health care services to an increasingly aging and frail population in a more cost-efficient manner. Beginning in the early 1990s, telephone-based health care initiatives began to multiply in the United States and Canada, partly in response to these imperatives. Existing under a variety of namesd“telephone case/care management”,5e7 “telephone-based disease management”,8,9 “telephonic” care,10,11 telephone “coaching/counselling/support” initiatives,12e14 among many othersd such programs can be defined as the ongoing provision of outpatient client and/or caregiver support (e.g., advice, education, care planning, assessment, advocacy, etc., but not including remote physiological monitoring), initiated primarily by a health care provider (e.g., registered nurse) over an extended period of time (e.g., weeks, months) primarily or exclusively by telephone. There is typically an expectation that the telephone intervention will have an effect on patient and health care system outcomes. For example, individuals with chronic heart failure who received frequent phone calls from a nurse over a 6-month period had reduced rates of heart failure hospitalization and fewer heart failure hospital days.15 However, periodic telephone contacts over 60 days actually increased the risk of hospital readmission in home care patients with diabetes or heart failure.16 And a recent evaluation of the

0197-4572/$ e see front matter Crown Copyright Ó 2015 Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.gerinurse.2014.11.004

112

R. Kelly, L. Godin / Geriatric Nursing 36 (2015) 111e119

“OwnHealth” telephone health coaching program in Birmingham, England, found that periodic telephone calls to individuals with chronic diseases did not lead to the expected reductions in hospital admissions or secondary care costs over 12 months.17 In general, the effect of telephone support programs have been assessed in a number of different outcomes, including health care costs,18e20 health-related quality of life,5,21,22 hospital admissions,13,23,24 clinical symptoms and self-care management6,25,26 and patient satisfaction.9,15,27 Overall, systematic reviews and meta-analyses have shown results to be mixed at best, and even contradictory in some instances.28e35 Taken together, previous telephone support programs have largely been directed at clients over a broad range of ages (e.g., from 18 years of age to 90 and older) with one or more of a relatively small set of chronic illnesses (e.g., heart failure, chronic obstructive pulmonary disease, congestive heart failure, diabetes) or conditions (e.g., substance abuse, obesity). Typically, clients had been recently discharged from hospital, and program length was generally short and fixed in duration (e.g., 3e6 months). In many instances, telephone contacts were frequent (e.g., weekly, biweekly), and telephone support was often supplemented with face-to-face visits from a health care professional, either in a clinic or at the client’s home. In general, the main objective of these telephone support initiatives was to follow up on clients who had recently experienced an acute episode of chronic illness, to ensure compliance with post-hospital care. Beginning in October 2010, the Fraser Health (FH) Authority in the lower mainland of BC introduced a “surveillance nurse” telephone support position as part of a larger Integrated Health Network implementation.36 Although the surveillance nurse position was inspired by previous telephone support programs, the initiative was unique in several key ways. In particular, the surveillance nurse supported only those clients in the FH home care program, consisting primarily of community-dwelling elderly individuals (e.g., on average, 80þ years of age) who were clinically and functionally stable but required health care supports to maintain independent living. One or more of a wide range of chronic diseases and/or functional deficits were present but were typically well managed by clients and their caregivers. Clients may have been receiving subsidized services (personal care and respite) or attending adult day programs. Other community resources (e.g., “meals-on-wheels”) may have been utilized as well. Contacts by the surveillance nurse were exclusively by telephone, and no remote physiological monitoring was involved. Scheduled calls

were generally less frequent (e.g., one call every one to six months), and a recent hospitalization was not required for clients to be admitted to the surveillance nurse’s caseload. In general, the main objective of the surveillance nurse initiative was to proactively monitor stable (but potentially morbid) clients’ well-being over time, in an effort to identify and address problematic issues before they became acute. A secondary objective of the initiative was to permit case managers to allocate greater time and effort to the more complex and unstable clients on their caseloads. The goal of the present study was to examine the effectiveness of the FH surveillance nurse telephone support initiative. Outcomes that were evaluated included delay to an event that terminated the home care episode (i.e., institutionalization or death), the rate of emergency room and hospitalization events during home care enrollment, and the rate of days in hospital during home care enrollment. A combined covariate-matching/propensity-scoring procedure37 was used to compare surveillance nurse clients to a highly similar group of clients that did not receive the surveillance nurse treatment. The surveillance nurse The conceptual foundation of the surveillance nurse position was broadly shaped by four evidence-based philosophies to health care, as shown in Table 1. Surveillance nurses are registered nurses with experience in the home care program. They are skilled communicators, educators, coaches and “care advocates”42 who are clinically astute and able to develop and maintain rapport and actively listen in order to assess clients’ conditions by telephone. Surveillance nurse caseload size is dependent on the number of home care clients in the community, but a single nurse can manage a caseload of up to 275 clients. Surveillance nurses participate in monthly teleconferences and annual face-to-face meetings to share experiences with and provide professional support and development to their colleagues. The surveillance nurse in the home care program Home care offices in FH can have up to 1300 clients each, depending on the size of the community. Home care clients must meet residency criteria and have either functional or cognitive deficits that require ongoing health care services to maintain independent living within the community.43 There must also be a desire on the part of clients and/or their caregivers to remain living at home, for as long as possible.

Table 1 Four conceptual foundations of the surveillance nurse position. Theory/Model

Description

Guided Care38

The integration of primary health care with recent innovations in chronic care (e.g., chronic disease self-management) to improve quality of care and thus optimize outcomes for patients with complex health care needs. In guided care, registered nurses complete a supplemental educational curriculum that gives them the special skills needed to practice. Each nurse then works in a practice with several primary care physicians and conducts clinical processes (assessment, care planning, monitoring, coaching, caregiver support) for a large caseload of multimorbid patients. Pertains to an individual’s sense of mastery and control over their own health. Self-efficacy beliefs are cognitions held by an individual which determine whether or not behavior change to support better health and well-being will be initiated and sustained by that individual. As individuals develop efficacy beliefs of greater strength, they become more likely to effect beneficial and lasting behavior change. Uses peers to support to educate people with various chronic diseases. This program was developed at Stanford University School of Medicine, and focuses on support for both disease- and non-disease specific actions (e.g., problem-solving, decision-making, dealing with emotions, etc.) and behavioral change. A framework that describes four pillars of a patient’s successful transition from hospital to home: (a) “red flag” warning signs, (b) medication reconciliation, (c) follow up with the primary care provider within a week, and (d) a client-held health record.

Self-Efficacy Theory39

Chronic Disease Self-Management Program40 Management of care transitions41

Transitional care refers to actions that ensure the coordination and continuity of health care as patients transfer between different levels and/or locations of service. A clinically-skilled “transitions coach” fosters communication amongst health care professionals and provides ongoing support to the client.

R. Kelly, L. Godin / Geriatric Nursing 36 (2015) 111e119

Upon an individual’s enrollment in the home care program a case manager conducts an initial face-to-face meeting with the client. The Resident Assessment Instrument for Home Care (RAI-HC)44 is used to document the client’s functioning with respect to health, cognition, psychosocial status and activities of daily living, and a plan of care is developed based on the assessment. The case manager authorizes subsidized personal care and/ or respite services (adult day program or in-facility respite) as needed. The service bundle and care plan are evaluated periodically and adjusted if necessary. During enrollment in the home care program, the client is reassessed periodically (e.g., every year) using the RAI-HC. At any time after enrollment, a client who is stable with respect to the care plan can be transferred to the surveillance nurse’s caseload. If care is transferred, the surveillance nurse assumes responsibility for the client. Prior to a telephone call, the surveillance nurse reviews the client’s care plan, service bundle, health concerns and any recent hospitalizations and emergency room registrations, as well as the client’s personal support systems and living situation. This review gives the surveillance nurse a summary of the client’s recent history and current circumstances. During the call the surveillance nurse assesses the client and provides education, coaching and information. Secondary screening instruments (e.g., a caregiver burden assessment) can be applied, if necessary. If the surveillance nurse identifies an “actionable” health-related issue, a Health Improvement Plan (HIP) is created for the client. The HIP formalizes the actions required to resolve the issue and indicates whether the actions are to be done by the client, caregiver or the surveillance nurse. The goal of the HIP is to empower the client and/or caregiver to gain control of and take responsibility for some of the issues that affect the client’s health. Progress on the plan is evaluated on a follow-up call. If needed, the surveillance nurse refers clients to allied health care professionals (occupational therapists, physiotherapists, social workers, dieticians and home care nurses) and community resources (e.g., meals-on-wheels, the Alzheimer’s Society, etc.), and adjusts clients’ care plans and subsidized services. The surveillance nurse might even collaborate with clients’ primary care providers (general or nurse practitioners) if necessary, although the preferred method is to coach the client to connect directly with the primary care provider. At the end of each telephone call, the surveillance nurse determines when the client would benefit from a follow-up call. Length and frequency of calls are tailored to the needs of the client, but call length typically varies between 15 and 30 min, and a client typically receives a call every one to six months. Clients are encouraged to contact the surveillance nurse between scheduled calls if their condition deteriorates, and to seek immediate medical care (i.e., consult with their primary care physician or attend an emergency room) if necessary. If in the surveillance nurse’s judgment a client’s situation or condition has worsened since the last call, the case manager is requested to meet with the client. If necessary, responsibility for the client’s care can be transferred back to the case manager. Clients are discharged from surveillance nurses’ caseloads if they are no longer receiving home care services and there is no active care planning or coaching. Method Data for this study came from all community-dwelling individuals who were aged 65 years and older when first admitted to the FH home care (HC) program in the 2-year period from July 1, 2012 to June 30, 2014. According to BC Ministry of Health policy, individuals referred to the HC program in FH must receive their first RAI-HC assessment shortly after admission.43 Thus, the electronic

113

system-generated creation date of individuals’ earliest completed RAI-HC assessment was adopted as a proxy for date of admission to the HC program. A total of 7664 community-dwelling individuals aged 65 years and older were first admitted to the HC program in this period. There were two reasons for using this particular 2-year period. First, a new electronic system for recording service utilization had been introduced in FH in mid-2012 and the previous electronic system was retired. Consequently, beginning the date frame at an earlier date would have required the integration of data from two disparate systems which would have greatly increased the complexity of obtaining an analytical data set. Second, although the surveillance nurse initiative actually began in September 2010, implementation was conducted over time as a phased rollout over communities in FH. Thus, it was only by mid-2012 that a substantial number of clients throughout FH had been admitted to a surveillance nurse’s caseload. Individuals in the study were monitored up to September 30, 2014 (the study cutoff date) to determine if they experienced a terminal eventdeither admission to an assisted living or nursing home facility or death. If one or more of these terminal events had occurred on or before the cutoff date, then the date of the earliest event was adopted as a proxy for date of discharge from the HC program, and the HC episode was defined as the number of days between admission and discharge. If a terminal event had not occurred by the cutoff date, then the cutoff date was used as the discharge date for determining the number of days in the HC episode. On the reasoning that individuals would require at least 90 days in the HC program to benefit from the services offered, a total of 1585 individuals with episodes of less than 90 days were dropped from the study, leaving a total of 6079 individuals in the data set.

Covariates The total amounts of FH subsidized home support (HS) hours and days of attendance at an adult day program (ADP) were obtained for all individuals. Utilization rates per week for these services were estimated by dividing individuals’ total amounts of service by the number of days between the first and the most recent dates of utilization, then multiplying by 7. So for example, if an individual used 150 h of HS service and there were 100 days between the first and the most recent dates of service use, then the HS utilization rate for that individual was estimated at 150/ 100  7 ¼ 10.5 h per week. Individuals’ ADP utilization rates were estimated in similar fashion except in units of days, rather than hours, per week. A total of 218 individuals had HS utilization rates of more than 20 h per week and a total of 74 remaining individuals had ADP utilization rates of more than 4 days per week. Upon closer inspection of the distribution of these two variables, these rates were deemed outliers and consequently these individuals were dropped from the study. And to eliminate the potentially confounding effects of respite services with the surveillance nurse treatment, an additional 104 individuals who had used respite were also dropped. Finally, a total of 97 individuals without a primary caregiver (G1eA) were dropped from the data set. Thus the final analytical data set consisted of N ¼ 5586 individuals. The following outcome scores were obtained from individuals’ first completed RAI-HC assessment (see Table 2): Cognitive Performance Scale (CPS); Method for Applying Priority Levels (MAPLe); Changes in Health, End-Stage Disease, Signs and Symptoms (CHESS); Activities of Daily Living Self Performance Hierarchy (ADL-SPH); Independent Activities of Daily Living Difficulty (IADL-D); Depression Rating Scale (DRS); Pain Scale (PS); and

114

R. Kelly, L. Godin / Geriatric Nursing 36 (2015) 111e119

Pressure Ulcer Risk Scale (PURS). In addition, the assessment provided two demographic factors, age and sex. Finally, the following covariates for individuals were computed from a combination of elements from their first completed RAI-HC assessments: Chronic Disease Score (i.e., CDS, ranging from 0 to 8, determined by the total number of the following eight diseases that were indicated on the assessment: cerebrovascular accident [J1a], congestive heart failure [J1b], coronary artery disease [J1c], hypertension [J1d], cancer [J1x], diabetes [J1y], emphysema [J1z], and renal failure [J1aa]); Dementia (“present” if either Alzheimer’s disease [J1g] or dementia other than Alzheimer’s [J1h] were indicated; “absent” if neither of these indicators were evident); Primary caregiver burden (“present” if the primary informal caregiver expressed an unwillingness to continue in the role [G2a], or dissatisfaction with support [G2b], or feelings of distress [G2c], or the individual and/or the primary caregiver felt the individual would be better off in a different living environment [O2b]; “absent” if none of these indicators were evident); and primary caregiver relationship (either “child/child-in-law”, “spouse” or “relative/friend” [G1fA]). For purposes of the analyses, sex, dementia, primary caregiver burden and primary caregiver relationship were considered to be nominal variables, and all other covariates were considered to be interval variables. Treatment and control conditions The “treatment” condition consisted of individuals in the final data set who had been admitted to a surveillance nurse’s caseload at any time during the date frame (nT ¼ 930), and the “control” condition consisted of all remaining individuals (nC0 ¼ 4656). Fig. 1 displays how the original data set was filtered to obtain the final analytical data set.

Propensity scoring and covariate matching For all individuals in the final data set, logistic regression was used to estimate a “propensity” score. A propensity score estimate represents the likelihood that an individual will be a member of the treatment condition, regardless of whether or not the individual actually receives the treatment. In other words, propensity score is the conditional probability that an individual receives the treatment given their values on a vector of known covariates. Covariates that are deemed essential to determining whether or not the treatment is received and the outcomes that are experienced are typically used to estimate the score. Propensity scores range between 0 and 1, with lower scores indicating a decreasing likelihood of treatment condition membership and higher scores indicate an increasing likelihood of inclusion in the treatment group. Interval covariates were used to estimate the propensity score: viz., age, CPS, MAPLe, CHESS, ADL-SPH, IADL-D, DRS, PS, PURS, CDS, HS hours per week and ADP days per week. Next, each individual in the treatment condition was matched to a single unique individual in the control condition, i.e., “one-to-one matching without replacement.” To be a match, the control individual first had to be an exact match to the treatment individual on all four nominal variables: viz., sex, dementia, primary caregiver burden and primary caregiver relationship. Second, given an exact match on the nominal variables, the control individual nearest in propensity score to the treatment individual was matched to that latter individual, i.e., “nearest neighbour” matching. Matching to individuals in the treatment condition proceeded from those with the highest propensity scores to those with the lowest scores. The matching procedure was implemented using the MatchIt package45 in the open-source R ver. 3.1.146 statistical programming language. Parameters used in the package were: method ¼ “nearest”; distance ¼ “logit”; replace ¼ FALSE; ratio ¼ 1; and m.order ¼ “largest”.

Outcomes Analysis plan The status of each individual at the end of the study and the number of days in the HC episode were used to estimate survival in the community. The total numbers of registrations in an FH emergency room, admissions in an FH hospital and days in an FH hospital for each individual during their HC episode were obtained. Days in the HC episode were used to determine rates of emergency room use, hospital admissions and total number of days in hospital. Rates were multiplied by 100 to produce rates per 100 days for each individual in the HC program.

Non-parametric (ManneWhitney U) tests were used to compare groups on covariates and outcomes for which there was substantial positive skew: viz., HS hours per week; ADP days per week; and averages and rates of emergency room registrations, hospital admissions and days in hospital per 100 days in the HC program. Conservative t-tests (equal group variances not assumed) were used to compare groups on interval covariates, and Pearson X2 tests were used to compare groups on nominal covariates. Finally,

Table 2 Covariates derived from the initial completed RAI-HC assessment. Interval covariates

Age (years) Range (lowest to highest risk)

Cognitive Performance Scale (CPS) Method for Applying Priority level (MAPLe) Changes in Health, End-Stage Disease, Signs and Symptoms (CHESS) Activities of Daily Living Self-Performance Hierarchy (ADL-SPH) Instrumental Activities of Daily Living Difficulty (IADL-D) Depression Rating Scale (DRS) Pain Scale (PS) Pressure Ulcer Risk Scale (PURS) Chronic Disease Score (CDS): Total number of the following eight diseases: Stroke (J1a), congestive heart failure (J1b), coronary artery disease (J1c), hypertension (J1d), cancer (J1x), diabetes (J1y), chronic obstructive pulmonary disorder (J1z), and renal failure (J1aa)

0e6 1e5 0e5 0e6 0e6 0e14 0e3 0e8 0e8

Nominal covariates

Values

Sex Presence of Alzheimer’s (J1g) or non-Alzheimers (J1h) dementia Caregiver is unable to continue (G2a), or is not satisfied with support received (G2b), or expresses feelings of distress (G2c), or client and/or caregiver feels client would be better off elsewhere (O2b) Primary caregiver relationship (G1fA)

Female, male Absent, present Absent, present Relative/friend, spouse, child/child-in-law

R. Kelly, L. Godin / Geriatric Nursing 36 (2015) 111e119

115

Fig. 1. Filtering conducted to obtain the final analytical data set. Note: HC ¼ home care, HS ¼ home support, ADP ¼ adult day program.

survival data were submitted to KaplaneMeier analyses and the ManteleCox log rank X2 statistic was evaluated. Due to the high number of statistical comparisons conducted in this study (viz., 53), Type 1 error rate was controlled by adopting an extreme significance level for each comparison, a ¼ 0.001. Hence, the study-wise error rate was a ¼ 1  (1  0.001)53 ¼ 0.052. Statistical analyses were done in IBM SPSS ver. 21.47 Results Before matching The first column of data in Table 3 displays descriptive statistics (means, standard deviations and percentages) on the covariates and outcomes for the 930 individuals in the treatment condition, the second data column displays the same descriptive statistics for the 4656 individuals in the control condition, and the third column of data displays the results of the comparison tests on all variables. For the interval covariates, there were no statistical differences between the two conditions in age, CHESS, PS, PURS, CDS and HS hours per week. However, individuals in the treatment condition had significantly lower CPS (t[1336] ¼ 6.53), MAPLe (t[1274] ¼ 6.85), ADL-SPH (t[1437] ¼ 4.07), IADL-D (t[1270] ¼ 5.53) and DRS (t[1441] ¼ 3.84) scores, and significantly higher ADP attendance (U ¼ 2,140,891) compared to individuals the control condition. In other words, conditions were statistically “unbalanced” with respect to these latter covariates: i.e., individuals in the treatment

condition were significantly more cognitively intact, at less risk of institutionalization, more independent in personal and independent activities of daily living, less depressed and attended ADP more frequently. For the nominal covariates, the two conditions did not differ (i.e., were “balanced”) in the percentages of females, individuals with dementia and relationship of the primary caregiver. However, primary caregivers of individuals in the treatment condition experienced significantly less burden than did primary caregivers of individuals in the control condition (X2(1) ¼ 40.97). With respect to the outcomes, significantly fewer individuals in the treatment than control condition experienced a terminal event, X2(3) ¼ 98.44. Thus not surprisingly, the HC episode was significantly longer in the treatment than control condition, t(1408) ¼ 12.60. Conditions did not differ in emergency room registrations or hospital admissions during the HC episode, although individuals in the control condition spent significantly more total days in hospital compared to individuals in the treatment condition, U ¼ 1,972,100. Individuals in the control condition had significantly higher rates per 100 days than individuals in the treatment condition for emergency room registrations (U ¼ 1,958,663), hospital admissions (U ¼ 1,931,221) and hospital days (U ¼ 1,930,542). Finally, the KaplaneMeier analysis disclosed significantly higher rates of survival in the community before a terminal event in the treatment compared to the control condition, X2(1) ¼ 123.05. Fig. 2 displays the proportions of individuals surviving in the community as a function of HC episode duration between day 90 and day 360,

116

R. Kelly, L. Godin / Geriatric Nursing 36 (2015) 111e119

Table 3 Descriptive and difference statistics. Before matching Treatment condition n ¼ 930 Interval covariates Age CPS MAPLe CHESS ADL-SPH IADL-D DRS PS PURS CDS

Control condition n ¼ 4656

Mean (SD) 83.2 1.3 3.2 1.0 0.7 4.0 1.1 1.2 0.9 2.8

(7.37) (1.14) (1.13) (0.96) (1.07) (1.62) (2.02) (1.02) (0.88) (2.21)

83.4 1.6 3.5 1.1 0.8 4.3 1.4 1.1 0.9 2.7

3.0 (3.96) 0.2 (0.44)

Difference test

t (df)

Mean (SD)

t (df)

(1350) (1336)* (1274)* (1373) (1437)* (1270)* (1441)* (1329) (1374) (1347)

U 3.7 (4.96) 0.1 (0.37)

Percentage

Female With dementia With primary caregiver burden Primary caregiver relationship Child/Child-in-law Spouse Relative/friend/none

66.5 17.4 31.0

63.0 18.3 42.2

62.5 26.9 10.6

59.0 27.5 13.6

Nominal outcomes

Percentage

Status at study cutoff date Admitted to assisted living Admitted to nursing home Died Censored

0.8 5.6 7.5 86.1

Interval outcomes

Mean (SD)

HC episode (days)

419.7 (168.74)

3.90 0.42 40.97 6.79

83.1 1.3 3.2 1.0 0.7 3.9 1.1 1.2 0.9 2.7

(7.53) (1.10) (1.11) (0.96) (1.10) (1.65) (1.95) (1.02) (0.90) (2.28)

Mean (SD)

2,140,891 1,999,750* X2 (df)

Nominal covariates

3.0 (4.28) 0.2 (0.44) Percentage

(1) (1) (1)* (2)

0.19 0.06 0.43 0.24 0.32 0.26 0.02 0.79 0.60 0.50

(1857) (1856) (1857) (1858) (1856) (1858) (1856) (1858) (1857) (1856)

U 394990 422716 X2 (df)

66.5 17.4 31.0

0.00 0.00 0.00 0.00

(1) (1) (1) (2)

62.5 26.9 10.6 X2 (df)

Percentage

98.44 (3)* 3.8 13.5 11.7 71.0

X2 (df) 60.67 (3)*

3.7 11.6 12.6 72.2

342.1 (184.23)

Mean (SD) Emergency room registrations Emergency room registration rate per 100 days Hospital admissions Hospital admission rate per 100 days Total hospital days Hospital day rate per 100 days

After matching Control condition n ¼ 930

0.75 6.53 6.85 1.31 4.07 5.53 3.84 2.77 0.43 0.41

(7.56) (1.15) (1.06) (1.01) (1.20) (1.51) (2.27) (1.03) (0.93) (2.26)

Mean (SD) HS hours per week ADP days per week

Difference test

t (df)

Mean (SD)

12.60 (1408)*

353.8 (185.83)

U

Mean (SD)

t (df) 8.01 (1841)* U

1.56 (2.31) 0.39 (0.56)

1.65 (2.65) 0.55 (0.87)

2,108,483 1,958,663*

1.66 (2.40) 0.55 (0.83)

418,083 392,854*

0.76 (1.34) 0.19 (0.35)

0.88 (1.40) 0.32 (0.53)

2,026,348 1,931,221*

0.86 (1.35) 0.31 (0.52)

408,789 393,636*

12.28 (38.84) 3.07 (8.57)

17.86 (39.03) 6.87 (15.41)

1,972,100* 1,930,542*

16.51 (34.81) 6.30 (14.40)

399,971 393,283*

CPS ¼ Cognitive Performance Scale, MAPLe ¼ Method for Applying Priority Levels, CHESS ¼ Change in Health, End-Stage Disease, Signs and Symptoms, ADL-SPH ¼ Activities of Daily Living Self Performance Hierarchy, IADL-D ¼ Independent Activities of Daily Living Difficulty Scale, DRS ¼ Depression Scale, PS ¼ Pain Scale, PURS ¼ Pressure Ulcer Risk Scale, CDS ¼ Chronic Disease Score, HS ¼ Home Support, ADP ¼ Adult Day Program, HC ¼ Home Care, SD ¼ standard deviation, df ¼ degrees of freedom. *p < 0.001.

in 30-day increments. At the 360-day mark of admission to the HC program, approximately 93% of individuals in the treatment condition were still in the program, whereas only about 73% of the control condition individuals were still in HC.

After matching The propensity-based covariate-matching procedure was effective in matching a unique individual in the control condition to each of the 930 individuals in the treatment condition. The mean propensity score for the 930 individuals in the treatment condition was 0.185 (standard deviation ¼ 0.060). Before matching, mean propensity score for the 4656 individuals in the control condition, 0.163 (standard deviation ¼ 0.055), was significantly lower, t(1267) ¼ 10.68. After matching, mean propensity score for the 930 matched individuals in the control condition, 0.184 (standard deviation ¼ 0.057), did not differ significantly from mean propensity score in the treatment condition, t(1855) ¼ 0.37.

The effectiveness in matching was confirmed by a comparison of the two conditions on the covariates. The fourth column of data in Table 3 displays descriptive statistics (means, standard deviations and percentages) on the covariates and outcomes for the 930 matched individuals in the control condition, and the final column of data displays the results of the tested comparisons to the 930 individuals in the treatment condition on all the variables. No significant differences were observed on any of the interval covariates. (Note that the percentages on the nominal covariates are exactly the same in the two conditions, because of the exact match required on these covariates during the matching procedure.) Thus, balance was evident on all the interval and nominal covariates. However, despite the balance on all these covariates, conditions continued to differ significantly on a number of the outcome variables. Compared to individuals in the treatment condition, individuals in the control condition experienced more terminal events, X2(3) ¼ 60.67, had shorter HC episodes, t(1841) ¼ 8.01, and had higher rates per 100 days of emergency room registrations,

R. Kelly, L. Godin / Geriatric Nursing 36 (2015) 111e119

117

Fig. 2. 360-day survival functions for treatment and control conditions before matching.

hospital admissions and hospital days (respective Us ¼ 392,854, 393,636 and 393,283). Conditions did not statistically differ in terms of absolute numbers of emergency room registrations, hospital admissions or total days in hospital. Finally, the KaplaneMeier analysis found the conditions also continued to differ significantly in terms of survival, X2(1) ¼ 78.03. Fig. 3 shows that at the 360-day mark of admission to the HC program only about 75% of the individuals in the control condition were still in HC compared to the approximately 93% of individuals in the treatment condition. Discussion Results of the present study revealed evidence of beneficial effects on several outcomes resulting from the surveillance nurse intervention in the FH home care program. Significant benefits were evident even after a propensity-based covariate-matching procedure eliminated differences between treatment and control conditions on a number of key covariates. Specifically, each individual in the treatment condition was matched to a single unique individual in the control condition who was identical on a set of four nominal covariates and whose propensity score, based on a set of twelve interval covariates, was nearest to that treated individual. Analyses indicated that balance on treatment and control conditions was achieved with respect to the sixteen covariates. Regardless, individuals in the treatment condition experienced better outcomes than their matched control counterparts, in terms of rate of emergency room registrations, hospital admissions and days in hospital per 100 days in the home care program, as well as survival in the home care program and the event that terminated the home care episode. It appears that the effect of the surveillance nurse on the rate outcomes was indirect, whereas the effect on duration of the home care episode and the survival outcome was direct. Specifically, although significant differences were observed between treatment and matched control individuals on rates of emergency room registrations, hospital admissions and days in hospital per 100 days in home care, significant differences were not observed in absolute numbers of emergency room registrations, hospital admissions and

days in hospital during the home care episode. Therefore, significant rate differences were likely due to differences between conditions in the overall duration of the home care episode. Consistent with this interpretation, the average number of days in the home care episode in the treatment condition, 419.7, was significantly greater than that in the control condition, 353.8, and the treatment survival function declined at a significantly slower rate than the control function. In short, conditions did not differ in terms of absolute numbers of emergency room registrations, hospital admissions or hospital days, but treatment individuals experienced those events in a home care episode of greater duration compared to individuals in the control condition. Although the surveillance nurse intervention had no effect on absolute numbers of emergency room registrations, hospital admissions and days in hospital, by extending the number of days in home care the intervention appeared to have an effect on the type of event that terminated the episode. Table 3 shows that the matched control condition had an overall greater percentage of individuals experiencing a terminal event compared to the treatment condition, as well as greater percentages in each category of terminal event. However, it could be argued that given enough time, the treatment condition would eventually reach the same percentages of individuals experiencing the various terminal events as in the control conditiondin essence, that the surveillance nurse merely postponed the terminal event in the treatment condition. To test this possibility, a further examination of data was done for only those individuals who experienced a terminal event to end the home care episode. In the treatment condition, 129 individuals experienced a terminal event that ended the home care episode: 7 (5.4%) were admitted to an assisted living facility, 52 (40.3%) were admitted to a nursing home and 70 (54.3%) died. In the matched control condition, 259 individuals experience a terminal event: 34 (13.1%) were admitted to an assisted living facility, 108 (41.7%) were admitted to a nursing home and 117 (45.2%) died. Perhaps the most notable difference between conditions is in the percent that died without having been admitted to an assisted living or nursing home facility: 54.3% in the treatment condition compared to 45.2% in the control condition. Evidently the surveillance nurse treatment increased the

118

R. Kelly, L. Godin / Geriatric Nursing 36 (2015) 111e119

Fig. 3. 360-day survival functions for treatment and control conditions after matching.

likelihood that a dying individual would experience their death in the community rather than in an assisted living or nursing home facility. Nevertheless, the average age (and standard deviation) at death for individuals who died while still living in the community, 84.9 years (7.37) and 85.8 years (7.37) respectively for treatment and control groups, did not differ significantly, t(145) ¼ 0.80. In other words, the surveillance nurse treatment did not simply postpone eventual death in the treatment group. Rather, it actually reduced the reliance on (and costs associated with) institutionalization for dying individuals. Several other telephone support evaluations have attempted to match treatment and control conditions on key covariates using methods that are simpler than propensity scoring.13,17 However, the present study is not the first propensity-based evaluation of a telephone support intervention. In a recent Japanese study, Akematsu and Masatsugu48 used a propensity score matching method with survey data to examine the effect of telecare on medical expenditures for individuals with chronic diseases. After successfully achieving balance through propensity score matching, the authors reported that the treatment group had lower medical expenditures and fewer treatment days for chronic diseases than the matched control group. In a similar study, Chumbler and colleagues49 implemented a form of propensity score matching in their study of a home/telehealth program for veterans with diabetes. They found the treatment improved individuals’ ability to receive appropriate timely care, thereby making more efficient use of health care resources. The present study replicates and extends these findings: The proactive surveillance nurse telephone support initiative had beneficial effects on a variety of health systems outcomes on an elderly home care population.

evident in individuals on the caseload for longer periods of time, was attenuated by a weaker or null effect, evident in individuals on the caseload for shorter periods of time. Unfortunately we were unable to test this possibility with the data used in this study. With the adoption of better record keeping practices since these data were gathered, future studies will be better equipped to explore the duration of surveillance nurse support on clients outcomes. There are also inherent shortcomings of the propensity-based covariate-matching procedure. Specifically, propensity and covariate matching can be done only for known and accurately measured covariates, and matching does not guarantee equivalence on unknown and poorly measured covariates. In other words, the success of the procedure is based on the so-called “strongly ignorable” assumption,37 i.e., all covariates related to assignment to conditions have been modeled by the propensity score. Short of the strongly ignorable assumption, systematic differences on unmeasured covariates can never be completely ruled out as the cause of outcome differences. To reduce the likelihood that unknown covariates were confounded with treatment self-selection, we used a broad set of covariates from multiple domainsddemographic, psychosocial, clinical and functionaldand we confirmed balance on those covariates after the matching process was complete. Rubin50,51 provides insightful discussions on how observational studies can approach randomized trials in terms of developing causal models, through the use of propensity score matching, and research has provided some support to this form of model-building.52,53 A recent textbook by Rosenbaum54 also offers an excellent and highly-readable source for the background and implementation of propensity scoring and covariate matching.

Limitations

Acknowledgments

Our data indicated whether or not an individual had been on a surveillance nurse’s caseload, but it did not contain information about the length of time on the caseload. There can be considerable variation over individuals in length of time on the caseload. In the present study it’s possible that a stronger surveillance nurse effect,

The authors wish to thank Irene Sheppard and Gloria Puurveen for their review of and comments on earlier drafts of this manuscript. The authors are also very grateful to the editor and anonymous reviewers for their helpful comments and suggestions on the initial submission of this manuscript.

R. Kelly, L. Godin / Geriatric Nursing 36 (2015) 111e119

References 1. Statistics Canada. Population by sex and age group, by province and territory. http://www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/demo31a-eng. htm; Published 2013 Accessed 09.04.14. 2. BCStats. British Columbia population Projections: 2013 to 2036. http://www. bcstats.gov.bc.ca/StatisticsBySubject/Demography/PopulationProjections.aspx; Published 2013 Accessed 09.04.14. 3. Ministry of Finance e The Province of British Columbia. Balanced Budget 2014. http://www.bcbudget.gov.bc.ca/2014/backgrounders/ 2014_backgrounder_1_fiscal_plan.pdf; Published 2014 Accessed 09.04.14. 4. Health Care in Canada, 2011: A Focus on Seniors and Aging. https://secure.cihi. ca/free_products/HCIC_2011_seniors_report_en.pdf; Published 2011 Accessed 09.04.14. 5. Riegel B, Carlson B, Glaser D, et al. Randomized controlled trial of telephone case management in Hispanics of Mexican origin with heart failure. J Card Fail. 2006;12(3):211e219. 6. Andrews G, Sunderland M. Telephone case management reduces both distress and psychiatric hospitalization. Aust N Z J Psychiatry. 2009;43:809e811. 7. Kim SE, Michalopoulos C, Kwong RM, et al. Telephone care management’s effectiveness in coordinating care for Medicaid beneficiaries in managed care: a randomized controlled study. Health Serv Res. 2013;48(5):1730e1749. 8. Dunagan WC, Littenberg B, Ewald GA, et al. Randomized trial of a nurseadministered, telephone-based disease management program for patients with heart failure. J Card Fail. 2005;11(5):358e365. 9. Azarmina P, Lewis J. Patient satisfaction with a nurse-led, telephone-based disease management service in Birmingham. J Telemed Telecare. 2007;13:3e4. 10. Galbreath AD, Krasuski RA, Smith B, et al. Long-term healthcare and cost outcomes of disease management in a large, randomized, community-based population with heart failure. Circulation. 2004;110(23):3518e3526. 11. Eldon Taylor C, Lopiccolo CJ, Eisdorfer C, et al. Best practices: reducing rehospitalization with telephonic targeted care management in a managed health care plan. Psychiatr Serv. 2005;56(6):652e654. 12. Afifi AA, Morisky DE, Kominski GF, et al. Impact of disease management on health care utilization: evidence from the “Florida: A Healthy State (FAHS)” Medicaid Program. Prev Med. 2007;44(6):547e553. 13. Lin WC, Chien HL, Willis G, et al. The effect of a telephone-based health coaching disease management program on Medicaid members with chronic conditions. Med Care. 2012;50(1):91e98. 14. Cleland JG, Louis AA, Rigby AS, et al. Noninvasive home telemonitoring for patients with heart failure at high risk of recurrent admission and death: the Trans-European Network-Home-Care Management System (TEN-HMS) study. J Am Coll Cardiol. 2005;45(10):1654e1664. 15. Riegel B, Carlson B, Kopp Z, et al. Effect of a standardized nurse casemanagement telephone intervention on resource use in patients with chronic heart failure. Arch Intern Med. 2002;162(6):705e712. 16. Bowles KH, Holland DE, Horowitz DA. A comparison of in-person home care, home care with telephone contact and home care with telemonitoring for disease management. J Telemed Telecare. 2009;15(7):344e350. 17. Steventon A, Tunkel S, Blunt I, et al. Effect of telephone health coaching (Birmingham OwnHealth) on hospital use and associated costs: cohort study with matched controls. BMJ. 2013;347:f4585. 18. Wong FK, Mok MP, Chan T, et al. Nurse follow-up of patients with diabetes: randomized controlled trial. J Adv Nurs. 2005;50(4):391e402. 19. Wennberg DE, Marr A, Lang L, et al. A randomized trial of a telephone caremanagement strategy. N Engl J Med. 2010;363:1245e1255. 20. Krumholz HM, Amatruda J, Smith GL, et al. Randomized trial of an education and support intervention to prevent readmission of patients with heart failure. J Am Coll Cardiol. 2002;39(1):83e89. 21. GESICA Investigators. Randomised trial of telephone intervention in chronic heart failure: DIAL trial. BMJ. 2005;331:425. 22. Bourbeau J, Julien M, Maltais F, et al. Reduction of hospital utilization in patients with chronic obstructive pulmonary disease: a disease-specific selfmanagement intervention. Arch Intern Med. 2003;163(5):585e591. 23. Reinius P, Johansson M, Fjellner A, et al. A telephone-based case-management intervention reduces healthcare utilization for frequent emergency department visitors. Eur J Emerg Med. 2013;20(5):327e334. 24. Wheeler EC, Waterhouse JK. Telephone interventions by nursing students: improving outcomes for heart failure patients in the community. J Community Health Nurs. 2006;23(3):137e146. 25. Maust DT, Mavandadi S, Benson A, et al. Telephone-based care management for older adults initiated on psychotropic medication. Int J Geriatr Psychiatry. 2013;28(4):s410es416. 26. Kim HS, Oh JA. Adherence to diabetes control recommendations: Impact of nurse telephone calls. J Adv Nurs. 2003;44(3):256e261.

119

27. Kirkman MS, Weinberger M, Landsman PB, et al. A telephone-delivered intervention for patients with NIDDM. Effect on coronary risk factors. Diabetes Care. 1994;17(8):840e846. 28. Balas EA, Jaffrey F, Kuperman GJ, et al. Electronic communication with patients: evaluation of distance medicine technology. J Am Med Assoc. 1997;278(2): 152e159. 29. McBride CM, Rimer BK. Using the telephone to improve health behavior and health service delivery. Patient Educ Couns. 1999;37:3e18. 30. Car J, Sheikh A. Telephone consultations. BMJ. 2003;326(7396):966e969. 31. Chaudhry SI, Phillips CO, Stewart SS, et al. Telemonitoring for patients with chronic heart failure: a systematic review. J Card Fail. 2007;13(1):56e62. 32. Tran K, Polisena J, Coyle D, et al. Home Telehealth for Chronic Disease Management [Technology Report Number 113]. Ottawa: Canadian Agency for Drugs and Technologies in Health; 2008. 33. Inglis SC, Clark RA, McAlister FA, et al. Which components of heart failure programmes are effective? A systematic review and meta-analysis of the outcomes of structured telephone support or telemonitoring as the primary component of chronic heart failure management in 8323 patients: Abridged Cochrane Review. Eur J Heart Fail. 2011;13(9):1028e1040. 34. Hutchison AJ, Breckon JD. A review of telephone coaching services for people with long-term conditions. J Telemed Telecare. 2011;17(8):451e458. 35. Desai AS. Home monitoring heart failure care does not improve patient outcomes: looking beyond telephone-based disease management. Circulation. 2012;125:828e836. 36. Park G, Miller D, Tien G, et al. Supporting frail seniors through a family physician and Home Health integrated care model in Fraser Health. Int J Integr Care. 2014;14(10):1e8. 37. Rosenbaum P, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrica. 1983;70(1):41e55. 38. Boult C, Karm L, Groves C. Improving chronic care: The “Guided Care” model. Perm J. 2008;12(1):50e54. 39. Bandura A. Self-efficacy: toward a unifying theory of behavioral change. Psychol Rev. 1977;84(2):191e215. 40. Lorig KR, Sobel DS, Stewart AL, et al. Evidence suggesting that a chronic disease self-management program can improve health status while reducing hospitalization: a randomized trial. Med Care. 1999;37(1):5e14. 41. Parry C, Coleman EA, Smith JD, et al. The care transitions intervention: a patient-centered approach to ensuring effective transfers between sites of geriatric care. Home Health Care Serv Q. 2003;22:1e14. 42. Alkema GE, Wilber KH, Shannon GR, et al. Reduced mortality: the unexpected impact of a telephone-based care management intervention for older adults in managed care. Health Serv Res. 2007;42(4):1632e1650. 43. Ministry of Health e The Province of British Columbia. Home and Community Care e Policy Manual. http://www2.gov.bc.ca/gov/topic.page? id¼8F569BDA913540DCAB75145DBB6070CE; Published 2014 Accessed 11.04.14. 44. Morris JN, Fries BE, Bernabei R, et al. RAI-Home Care (RAI-HC) User’s Manual, Canadian Version. Washington, DC: interRAI; 2010. 45. Ho DE, Imai K, King G, Stuart EA. MatchIt: nonparametric preprocessing for parametric causal inference. J Stat Softw. 2011;42(8):1e28. URL: http://www. jstatsoft.org/v42/i08/. 46. R Development Core Team. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, ISBN 3900051-07-0; 2008. URL: http://www.R-project.org; 2008. 47. IBM Corp. IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp; Released 2012. 48. Akematsu Y, Tsuji M. Measuring the effect of telecare on medical expenditures without bias using the propensity score matching method. Telemed J E Health. 2012;18(10):743e747. 49. Chumbler NR, Vogel WB, Garel M, Qin H, Kobb R, Ryan P. Health services utilization of a care coordination/home-telehealth program for veterans with diabetes: a matched cohort study. J Ambul Care Manage. 2005;28(3):230e240. 50. Rubin DB. The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials. Stat Med. 2007;26: 20e36. 51. Rubin DB. Using propensity scores to help design observational studies: application to tobacco litigation. Health Serv Outcomes Res Methodol. 2001;2: 169e188. 52. Luellen JK, Shadish WR, Clark MH. Propensity scores: an introduction and experimental test. Eval Rev. 2005;29(6):530e558. 53. Shadish WR, Clark MH, Steiner PM. Can nonrandomized experiments yield accurate answers? A randomized experiment comparing random and nonrandom assignments. J Am Stat Assoc. 2008;103(484):1334e1350. 54. Rosenbaum PR. Design of Observational Studies. Springer Series in Statistics. New York; 2010.