Research in Social and Administrative Pharmacy 11 (2015) 639–650

Original Research

Factors influencing community pharmacists’ likelihood to ask medication monitoring questions: A factorial survey Matthew J. Witry, Pharm.D., Ph.D.*, William R. Doucette, Ph.D. University of Iowa College of Pharmacy, Iowa City, USA

Abstract Background: Community pharmacists are well positioned to identify and resolve medication related problems associated with chronic medication use during prescription dispensing, a process referred to as medication monitoring. Pharmacists need feedback about patient medication experiences to engage in effective monitoring, but the pharmacist’s decision making process for when to ask questions to solicit this information from patients has not been established. Objectives: Identify significant factors contributing to a community pharmacist’s likelihood to ask medication monitoring questions at the time of refill. Methods: A factorial survey approach was used to test the effect of several pharmacist, patient, environment, drug, and past interaction factors (the domains of the Health Collaboration Model) on a pharmacist’s selfreported likelihood to ask non-adherence, side effect, and effectiveness monitoring questions for 5 randomly populated refill prescription dispensing vignettes. Surveys containing the vignettes, demographic items, and a new medication monitoring attitude measure were mailed to 599 community pharmacists. Hierarchical linear regression was used to test the independent effects of the vignette and pharmacist factors. Results: There were 254 (42.4%) returned and usable surveys. The hierarchical linear regression models showed that adherence questioning was driven more by the vignette characteristics whereas side effect and effectiveness questioning were more driven by the pharmacist. Overall, warfarin and hydrocodone were seen as more question-worthy than fluoxetine or metoprolol. The number of additional persons waiting in the pharmacy decreased, and more days late increased the likelihood of asking the three monitoring questions. An exception was hydrocodone where early fills prompted question asking. For side effect and effectiveness questioning, being short-staffed and the prescription previously being filled more times decreased question asking likelihood. Discussion: Factorial surveys are a useful approach to independently measuring the impact of respondent and contextual variables on pharmacist judgments. Reactions to the vignettes demonstrated that multiple factors go into a pharmacist’s mental model when deciding to ask a question at the time of refill. The

Disclosure: The authors declare no conflicts of interest or financial interests in any product or service mentioned in this article, including grants, employment, gifts, stock holdings, or honoraria. Funding: Dr. Witry was supported in part by a predoctoral fellowship from the American Foundation for Pharmaceutical Education; however, they had no involvement in the research. * Corresponding author. Department of Pharmacy Practice and Science, University of Iowa College of Pharmacy, 115 S. Grand Ave. S515 PHAR, Iowa City, IA 52242, USA. Tel.: þ1 319 335 8763; fax: þ1 319 353 5646. E-mail address: [email protected] (M.J. Witry). 1551-7411/$ - see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.sapharm.2014.11.007

640

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650

lateness of a refill prescription was a significant cue to question asking. Pharmacies can ensure late refill information is reaching pharmacists as a means to increase in medication monitoring. Pharmacies also can design work environments and workflows conducive to question asking and prompt pharmacists to address potentially under-discussed medications. Ó 2015 Elsevier Inc. All rights reserved. Keywords: Pharmacist; Counseling; Factorial survey; Community pharmacy

Introduction Medications are the most commonly used therapy in managing chronic disease. While often effective, medications can be associated with problems which are difficult to anticipate and a profound effect on health in the U.S. An estimated 4.5 million ambulatory care visits1 and 100 thousand deaths2 are attributed to medication misadventures annually. Historically, health care system improvements designed to support appropriate prescribing (e.g. condition-specific order sets) have received the most attention.3 Recent evidence, however, suggests the majority of medication related problems are not the result of inappropriate prescribing, but rather from the patient experiencing an unfortunate side effect, not experiencing the intended benefit, or not adhering to the therapy.2,4–8 Ongoing medication monitoring is required to identify and manage these types of problems. Tackling non-adherence alone has the potential to save the U.S. health care system $290 billion annually through improved chronic disease management.9 Community pharmacists are well positioned to monitor the ongoing use of medications because patients present to community pharmacies to obtain 1.5 billion prescription refills annually.10 Community pharmacists have demonstrated success in specialized medication monitoring services for conditions such as HIV11 and dyslipidemia.12 Most community pharmacists, however, are not engaged in specialty services, rather they spend the majority of their time in a dispensing role.13 These pharmacists, at least partially responding to time and staffing constraints, gravitate toward evaluating the appropriateness of, and counseling on, new prescriptions more so than counseling on, and evaluating patient experiences with refill medications.13–17 When dispensing pharmacists do engage in medication monitoring of refill prescriptions, it appears to be on an ad hoc, rather than a systematic basis.14,15,18–20 Pharmacists may be reacting to specific stimuli in deciding when to interact with a patient regarding

an ongoing refill of a maintenance medication. There is a need to identify what factors influence this process to better understand how pharmacists approach medication monitoring in the refill dispensing process so that interventions and education can be designed to support pharmacists in their monitoring efforts. The Health Collaboration Model provides a suitable framework for examining medication monitoring.21 This structure-process-outcome model, grounded in the patient centered communication literature, focuses on how patient, provider (pharmacist), drug, environment, and relationship factors contribute to quality medication monitoring. Patient centered medication monitoring involves engaging the patient by asking questions about medication experiences and preferences.15,16,22 Ultimately, the model purports improved monitoring leads to better identification and resolution of medication related problems on route to improved patient outcomes. The objective of this analysis was to identify significant factors contributing to a community pharmacist’s likelihood to ask medication monitoring questions at the time of refill. To accomplish this objective, we used a factorial survey approach to test the influence of the following structural factors of the Health Collaboration Model (Fig. 1): pharmacist factors (gender, degree, position, hours worked, monitoring attitudes); patient factors (gender, age, refill days date); environmental factors (pharmacy type, busyness, general and situational staffing levels, external work attitudes); drug factors (medication, times the medication was previously filled); and past interaction (relationship), influences the likelihood of asking monitoring questions at the time of refill. Materials and methods Factorial survey approach This analysis is part of a larger, sequential exploratory mixed methods23 study of community pharmacist medication monitoring. This report details the use of a quasi-experimental

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650

641

Fig. 1. Conceptual model and hypotheses based on health collaboration model21 and qualitative interviews.20

factorial survey approach to test the 5 study hypotheses of influences on a pharmacist’s decision to ask questions at the time of refill. A factorial survey uses a series of vignettes to assess an individual’s beliefs, decision making process, or judgments.24 The vignettes share a common structure and are populated with random values for a set of embedded variables (e.g. the patient’s age). This process creates orthogonal (uncorrelated) situations of which the individual records a response using a scale. The factorial survey approach has several advantages over other methods. In a traditional static vignette study, one can only speculate what explains the responses for the vignettes that the researcher created given his or her pre-existing assumptions of what would be important to put forward in each vignette. A factorial survey yields the independent effects of each included variable.24 In this way, factorial surveys are similar to discrete choice experiments and conjoint analyses.25 Factorial surveys also have certain advantages over observational secret shopper studies26 (a mainstay of the pharmacist counseling research) in that they can be conducted when using a confederate would not be feasible such as with the present focus on refill prescriptions. In addition, the simulated nature of the study allows the investigation of more extreme scenarios in

greater numbers taking less time be feasible in a specific behavior

and with more variation while for those involved than would structured observation where a is observed.

Vignette skeleton and variables For the factorial vignettes, a skeleton was developed based on the domains of the Health Collaboration Model21 and context provided by the coding and analysis of a series of semistructured qualitative interviews with community pharmacists20 and read as follows: You are verifying a refill prescription for a ! AGEO year old !GENDERO !FAMILIARITYO. The prescription is for !DRUGO. The prescription previously has been filled !FILLED TIMESO and is !EARLY/LATEO. The patient is waiting, there are !WAITINGO waiting in the pharmacy, and the pharmacy is !STAFFEDO. Vignette variable details are found in Table 1 and the mapping of the variables onto the domains of the Health Collaboration Model are found in Fig. 1. Patient age and gender were included because gender differences have been reported in the physicianpatient communication literature27 and pharmacists may give extra attention to older patients as they may be more sensitive to medication effects. The

642

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650

Table 1 Level-1 (Vignette) and Level-2 (Pharmacist) variables for factorial survey analysis Level-1 (Vignette) Variable

Conditions

Integer between 35–89a Male (1) Female (0) Days late Number of days late or early. Integer between 3 and 14 (including on-time ¼ 0) Other waiting Number of persons waiting. Integer between 0 and 4 Short staffed Short-staffed (1) Staffed as usual (0) Drug Fluoxetine Fluoxetine 40 mg, take one capsule daily, #30 Drug Hydrocodone Hydrocodone/APAP 5/325, take two tablets four times a day as needed for back pain, #120 Drug Metoprolol Metoprolol succinate 100 mg, take one tablet daily, #30 Drug Warfarin (reference) Warfarin 7.5 mg, take one tablet daily, #30 Filled Number of times previously filled. Integer between 1 and 10 times Familiar Whom you know by name (1) With whom you are not very familiar (0) Pt age Pt male

Level-2 (Pharmacist) Variable

Responses

RPh male

Male (1) Female (0) Yes (1) No (0) Staff (1) Manager/owner (0) Independent pharmacy (1) Chain/grocery/mass merchandiser pharmacy (0) Continuousa number of hours worked per week Continuous number of other pharmacists usually working Internal MMAM mean (possible range: 1–6)a External MMAM meanb (Possible range: 1–6)a

RPh PharmD RPh staff RPh indep Hours/week Other pharmacists Internal MMAM External MMAM

MMAM is Medication Monitoring Attitude Measure. a Grand mean centered. b Measure is reverse coded so that higher numbers are associated with an environment more conducive to medication monitoring.

earliness or lateness of the prescription was attributed to the patient and included because this is one of the only pieces of monitoring information universally available to community pharmacists through their computer system. For the environment, the number of other patients waiting was included because a sense of busyness is a common phenomena pressuring both pharmacists and patients.19,20 Whether or not the pharmacy was short staffed was included because discretionary refill counseling may decrease when pharmacists or technicians are missing from the workflow and the technical and legally mandated elements of dispensing demand immediate attention.20 The four drugs were selected because they were mentioned during the interviews in a way that suggested pharmacists are considering

the characteristics of certain medications when making judgments about when to talk with patients.20 The drugs and doses included: warfarin 7.5 mg (risk of bleeding, stroke), metoprolol succinate 100 mg (common but can have side effects and associated with established blood pressure goals), hydrocodone 5/325 (controlled substance with potential for abuse), fluoxetine 40 mg (common but pharmacists may not be as comfortable monitoring based on stigma associated with depression or uncertainty assessing effectiveness).28,29 The number of times the medication had previously been filled was included because some pharmacists had shared that newer medications garnered more attention because they believed problems were more likely.20 Lastly, familiarity was included

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650

because in the interviews, some pharmacists reported engaging more with patients with whom a prior relationship had been developed.20 Each survey contained 5 randomly populated vignettes. Vignettes were constructed using the random number feature in Microsoft Excel and the mail merge feature of Microsoft Word (Microsoft, Redmond WA), although other approaches to making randomly populated vignettes are possible. The total number of possible vignettes in the vignette universe was 1,710,720, of which only a subset were sampled as the variable distribution was random. For each vignette, the pharmacist was instructed to judge his or her perceived likelihood on a scale of 0–10 (0 ¼ definitely would not ask, 10 ¼ definitely would ask) for each of three medication monitoring questions: Ask the patient questions to identify possible reasons for non-adherence; Ask the patient questions to find out if s/he might be experiencing side effects; Ask the patient questions to find out about the effectiveness of the medication. These 3 question types were based on published medication monitoring focus areas.3 An example vignette and response is below. You are verifying a refill prescription for a 66 year old female of whom you know by name. The prescription is for fluoxetine 40 mg, take one capsule daily, #30. The prescription previously has been filled 2 times and is 4 days late. The patient is waiting, there is one other person waiting in the pharmacy, and the pharmacy is staffed as usual.

Likelihood scale: Definitely would not ask – 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 – Definitely would ask. Ask the patient questions to identify possible reasons for non-adherence ______ Ask the patient questions to find out if she might be experiencing side effects ______ Ask the patient questions to find out about the effectiveness of the medication ______ Survey administration A mailed survey was used as the method of data collection, in part because email addresses were not readily available and because the available online survey tools did not allow for unique surveys. A draft survey was reviewed by 8 pharmacists, 6 providing written comments, and 2 engaging in a verbal cognitive interview process. Small formatting and content changes were made based on this feedback. A single mailing of 50 pilot surveys was used to gauge response rate and response variation. Based on a response rate of

643

24% in this pilot, the length was shortened by decreasing the number of vignettes from 6 to 5, omitting 6 Likert items deemed duplicative, and formatting changes. For the main survey mailing, a 4-contact Dillman approach was utilized consisting of a pre-notification letter, an initial survey packet, a reminder postcard, and a second survey packet for non-responders.30 Surveys were confidential and could be made anonymous by tearing off the survey ID number as described in the cover letter. Unique surveys were mailed to 599 community pharmacists randomly selected from a list of the 1861 community pharmacists registered to practice in Iowa provided by the state board of pharmacy. Surveys were mailed in June and July, 2013. The study was approved by the Institutional Review Board of the University of Iowa. Survey items To describe the sample and test pharmacist and environment factors, pharmacists responded to a set of demographic and workplace items. For pharmacist factors, gender was selected as a general demographic variable. Degree was included because the PharmD curriculum has a greater emphasis on patient counseling than the B.S. Pharm curriculum. Whether a pharmacist was a staff or manager/owner was included because the latter may be more invested in developing relationships with patients given their financial stake in the pharmacy. Hours worked per week was included because pharmacists only working a small number of hours may not know the patients as well or be as comfortable with the workflow resulting in fewer interactions. For environmental factors, working at an independent pharmacy was included because these pharmacies may place greater emphasis on patient interaction as part of their business model. The average number of other pharmacists working was included because pharmacists usually working alone may have adopted an overall orientation where they are less likely to engage patients due to workflow. The pharmacy being short staffed was included because some respondents in the qualitative study reported being more likely to counsel on refills when another pharmacist was working.20 Two attitudinal orientation scales related to medication monitoring also were included, each using a 6-point scale (1 ¼ strongly disagree, 6 ¼ strongly agree). The validation of the Medication Monitoring Attitude Measure (MMAM)

644

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650

are detailed elsewhere.31 Seven of these items comprise the “internal” MMAM and contain items about pharmacist motivation, responsibility, and role perception of medication monitoring and were included as a pharmacist factor. Eight items comprised the “external” MMAM and focused on the environmental factors of busyness and perceived patient interest in medication monitoring for the clientele of the pharmacy. This measure of barriers to monitoring was reverse coded in the present analysis so that higher values would be associated with greater monitoring likelihood for both scales. Surveys were entered into Microsoft Excel (Microsoft, Redmond WA). Separate databases were constructed for vignette level and pharmacist level data, linked by a survey identifier. Descriptive analyses were performed using SPSS v.21 (IBM, Armonk NY) and HLM v.7 (SSI, Skokie IL) was used for the factorial survey analysis. Hierarchical linear regression (also referred to as random intercepts linear modeling or multi-level modeling) is necessary when analyzing factorial surveys because of theorized correlation between the individual’s responses.32,33 Using standard multiple regression would violate the assumption of independence.33 With factorial surveys, the characteristics of the respondent can be modeled on the dependent variables independent from the vignette characteristics. Respondent characteristics and vignette level variables are simultaneously estimated based on their own random variation. Output is interpreted as would any multiple regression – a 1 unit increase in X is associated with a B unit increase in Y, controlling for other variables. The following analysis procedure was used to test the relevance of variables entered into the models. First, a null model was tested to produce intraclass correlations (ICC) for each of the 3 question types to determine the proportion to which respondent level variation contributes to variation in the question asking likelihoods.34 Both restricted maximum likelihood and maximum likelihood estimation (MLE) produced the same coefficients which supports model building by adding variable groups and testing using MLE for improvements in model fit based on statistically significant log-likelihood tests (P ! 0.05).34 The first model added vignette characteristics, the second added an interaction term between “days late” and the “hydrocodone” prescription based on the proposition that pharmacists would react more strongly to an early controlled substance whereas they would react more to a late chronic maintenance

medication. Next, pharmacist workplace and demographic variables were added, followed by the two (internal and external) medication monitoring attitude measures (MMAM). Robust standard error estimates and standard estimates were compared to support that the modeling did not meaningfully violate hierarchical linear regression assumptions.34 Since the vignettes are orthogonal, and the dependent variables have a restricted range, unmeasured variables are expected to manifest as unexplained error rather than biased estimates. The values associated with the vignettes were tested to confirm an orthogonal structure for both the mailed and returned surveys.

Results For the main survey mailing, 599 surveys were mailed and none were returned undeliverable. Of these, 279 were returned (46.6%). Twenty five surveys were removed from the final data set. These included 17 from pharmacists who were retired or not practicing in community pharmacy, 5 returned blank or did not want to participate, and 2 showed clear signs of satisficing. This left a total usable sample of 254 pharmacists (42.4%). The majority of survey respondents (62.2%) were female, 59.1% had a BS Pharm degree, and 65.4% were staff pharmacists (Table 2). Intraclass correlations (ICC) for the three null models associated with question topic were calculated to determine if there was meaningful grouping at the respondent level. All three ICC’s were statistically significant, supporting the need for multi-level modeling.32 Specifically, the Ask Non-adherence null model ICC was 0.367, the Ask Side Effect null model was 0.566, and the Ask Effectiveness null model ICC was 0.614. ICCs !0.500 suggest the vignette characteristics drive the majority of the variation and ICCs O0.500 suggest the respondent drives the majority of the variation in the dependent variable.32,34 For the model building process, adding the vignette characteristics, pharmacist attitudes about monitoring, and an interaction term between hydrocodone and days late, were all significant improvements in model fit (P ! 0.01). The set of pharmacist demographic factors which included degree type and pharmacy setting showed weaker improvements (P ! 0.05). For the model of the likelihood of asking a non-adherence question in response to a random

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650 Table 2 Characteristics of survey respondents Variable Gender (male) PharmD degree Staff pharmacist Manager Owner Independent pharmacy Chain pharmacy Grocery pharmacy Mass merchandiser pharmacy

Yes 98 105 166 80 23 85 116 41 14

%

N

37.8 41.3 65.4 31.5 9.1 33.5 45.7 16.1 5.5

254 254 254 254 254 254 254 254 254

Variable

Mean

Std. dev

N

Year of first licensure Hours worked per week Typical number of other pharmacists Internal MMAM External MMAMa

1992.78 36.05 0.82

13.11 11.55 0.84

206 253 253

4.62 3.87

0.68 0.88

254 254

Medication Monitoring Attitude Measure (MMAM) scale is 1 ¼ Strongly disagree, 2 ¼ moderately disagree, 3 ¼ slightly disagree, 4 ¼ slightly agree, 5 ¼ moderately agree, 6 ¼ strongly agree. a Measure is reverse coded so that higher numbers are associated with an environment more conducive to medication monitoring.

vignette (Table 3) male pharmacists, those with BS Pharm degrees, and pharmacists with higher external and internal MMAM means reported higher likelihoods independent of vignette characteristics. Pharmacists also were more likely to ask questions about non-adherence if the patient was female and if fewer other patients were waiting. Pharmacists were equally likely to ask about non-adherence for hydrocodone and warfarin. Fluoxetine and metoprolol were significantly less likely to precipitate questions about nonadherence compared to warfarin. For the model of the likelihood of asking a side effect question (Table 3), the only significant pharmacist factors were internal and external MMAM means. For vignette factors, warfarin and hydrocodone again were equally likely to precipitate questioning. Metoprolol and fluoxetine had a significantly lower likelihood than warfarin, the reference medication. Prescriptions that had been filled fewer times were significantly associated with a greater likelihood, as was fewer patients waiting, and being “staffed as usual.” A late prescription for metoprolol, fluoxetine or warfarin was positively associated with asking about side effects, whereas an early prescription

645

for hydrocodone was associated with asking side effect questions, as shown by the negative interaction term of hydrocodone and days late. For the model of the likelihood of asking an effectiveness question (Table 3), the only significant pharmacist factors were internal and external MMAM means. At the vignette level, hydrocodone was significantly positively associated with effectiveness question asking likelihood compared to warfarin. Warfarin, fluoxetine, and metoprolol were not significantly different. Prescriptions that had been filled fewer times, or were late (warfarin, fluoxetine, and metoprolol) or early (hydrocodone) were positively associated with increased likelihood to ask effectiveness questions. Having other patients waiting and being short staffed were negatively associated with asking about effectiveness. Discussion The objective of the present study was to identify significant pharmacist and contextual factors which contribute to pharmacists’ decisions to ask medication monitoring questions at the time of refill. Overall, both the pharmacist and vignette characteristics were important to question asking likelihood, but to different degrees. Non-adherence question asking was more driven by the vignette characteristics whereas side effect and effectiveness questions were more driven by pharmacist characteristics. This pattern suggests that side effect and effectiveness questioning are based more on pharmacist motivation and environment, and nonadherence questioning is based more on the context, such as the number of days late of the prescription. These findings reinforce the basic premise that pharmacists were cognitively evaluating these hypothetical scenarios. A similar interpretation was proposed in a previous static vignette study of pharmacists evaluating dispensing scenarios.35 Pharmacist factors The pharmacists’ medication monitoring attitude were significant predictors of the likelihood to ask the three monitoring questions, whereas demographic or workplace factors were not (Table 3). This suggests there are pharmacists across settings and roles that are oriented to medication monitoring, although the ability to identify small significant differences in pharmacist factors was hindered by the limited number of vignettes given to each pharmacist. This is encouraging because pharmacists are being advanced by the health care community to help improve patient outcomes.3

646

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650

Table 3 Full model comparison for each question asking likelihood responses Ask non-adherence Vignette factors (Level 1) n ¼ 1270 0.00265 (0.00367) Pt agea Pt – male 0.298* (0.118) Days late 0.327** (0.0187) Other waiting 0.130** (0.0467) Short staffed 0.0780 (0.122) Drug- fluoxetine 1.288** (0.179) Drug- hydrocodone 0.317 (0.260) Drug- metoprolol 1.015** (0.179) Drug- warfarin (reference) – Times previously filled 0.0219 (0.0205) Familiarity- know by name 0.0244 (0.124) (Hydrocodone)* (days late) 0.385** (0.0322) Pharmacist Factors (Level 2) n ¼ 254 Intercept 2.994** (0.330) RPh male 0.576* (0.246) RPh PharmD 0.637* (0.250) RPh staff 0.136 (0.259) RPh independent 0.247 (0.271) Hours/weeka 0.00801 (0.0129) Other pharmacists 0.0711 (0.174) Internal MMAMb 0.713** (0.176) External MMAMa,b 0.569** (0.153) -2LL 5663.50**

Ask side effect

Ask effectiveness

0.00335 0.0369 0.165** 0.150** 0.382** 0.629** 0.0357 0.354* – 0.0939** 0.000919 0.158**

0.000156 0.0554 0.142** 0.163** 0.356** 0.0251 0.770** 0.141 – 0.0720** 0.0879 0.178**

(0.00356) (0.100) (0.0141) (0.0413) (0.114) (0.182) (0.255) (0.193)

4.188** 0.376 0.293 0.437 0.138 0.00814 0.121 1.142** 0.731** 5561.24**

(0.398) (0.311) (0.315) (0.316) (0.301) (0.0146) (0.179) (0.256) (0.216)

4.189** 0.402 0.461 0.0627 0.00316 0.000269 0.000454 1.159** 0.722** 5452.13**

(0.00332) (0.106) (0.0119) (0.0374) (0.107) (0.150) (0.200) (0.154) (0.0186) (0.107) (0.0235) (0.374) (0.285) (0.287) (0.307) (0.279) (0.0131) (0.158) (0.211) (0.174)

(0.0210) (0.108) (0.0271)

Values are linear regression coefficients (with robust standard errors) based on responses using an 11 point likelihood scale, 0 ¼ Definitely would not ask to 10 Definitely would ask for each of the three monitoring question types. *P ! 0.05, **P ! 0.01. Medication Monitoring Attitude Measure (MMAM) scale is 1 ¼ Strongly disagree, 2 ¼ moderately disagree, 3 ¼ slightly disagree, 4 ¼ slightly agree, 5 ¼ moderately agree, 6 ¼ strongly agree. a Measure is reverse coded so that higher numbers are associated with an environment more conducive to medication monitoring. b Grand mean centered.

Patient factors Female patients were more likely to be questioned about non-adherence than male patients in the vignettes (Table 3). Research on patient gender and health communication are mixed, but some have shown that women ask more questions, make sure they understand the physician as evident by paraphrasing physician instructions, and are more likely to engage in tension release (laughter) compared to males.36 These patterns also may be present for patients visiting community pharmacies and may translate to pharmacists having experience with female patients being more receptive to discussing non-adherence than male patients. The number of days late (or early in the case of hydrocodone) was significant for all three

question types (Table 3). The number of days late was the only evidence of a patient’s medication use experience with which the pharmacist could include in their mental model for question asking likelihood. Days late is a common means for pharmacists to identify non-adherence and appears to be an important activator for pharmacists engaging in medication monitoring. However, this information may not be readily available to the pharmacist in some workflows.20 It is encouraging that when presented with the number of days late, the pharmacists perceived a greater likelihood of asking monitoring questions. It has been observed, however, that pharmacists may limit their adherence discussion to logistical reasons for non-adherence,16,20,37 such as asking about dose changes or assuming the patient has an extra supply, rather than exploring

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650

beliefs and experiences, which have been shown to be responsible for a significant proportion of nonadherence.38 Environmental factors The number of patients waiting was important to pharmacist question asking, which corroborates other community pharmacy studies reporting barriers to counseling.19,39–41 Bitner suggests the images associated with busyness influence how customers (or patients) perceive the level of service provided by a firm.42 If pharmacies want to engage in more medication monitoring, it may be advisable to better manage this image, particularly as Pharmacists look to engage with patients as part of value based health care initiatives and programs that reward pharmacists based on quality.43 The pharmacy being short staffed was an important deterrent for side effect and effectiveness questions, but not for non-adherence (Table 3). This supports the proposition that side effect and effectiveness questions are more discretionary and subject to the pharmacist having adequate staff support. This is in contrast to non-adherence questioning which may be more robust to staffing levels because there is objective evidence of the need to interact based on number of days late. The pattern differences for these environmental factors also display the methodological ability of the factorial survey method to uncover tendencies that may be socially undesirable or difficult to access using a direct question approach. Drug factors Pharmacists were most likely to ask nonadherence and side effect questions related to warfarin and hydrocodone (Table 3). The importance of warfarin likely was due to the potential seriousness of misusing this medication as far as increased risk of stroke or bleeding. Interestingly, after adding an interaction term between hydrocodone and days late, pharmacists also found refilling hydrocodone early to be at this higher level of concern. The importance for hydrocodone may partially concern overuse. Less important was asking questions about metoprolol and fluoxetine. These medications may be perceived as less immediately risky or less actionable given the interventions with which the pharmacist is comfortable. This lesser emphasis may not be ideal because long term, safe, and effective use of these

647

medications has a significant impact on preventing heart attacks and strokes44 (metoprolol), and quality of life and improved self-management of chronic diseases45 (fluoxetine). For asking about effectiveness, respondents were most likely to ask about hydrocodone. The prescription for hydrocodone was for a fairly large number of tablets (#120) to be used as needed. The pharmacists may have expected this supply to last longer and may be interested in asking the patient questions about his or her pain level. Alternatively, respondents may have been concerned about prescription drug abuse or diversion and such investigative questions fell under the respondent’s conceptualization of “effectiveness.” One barrier to effectiveness monitoring for chronic medications may be a lack of readily available information about treatment goals. Pharmacists have reported that a lack of patient information makes them less likely to intervene based on an assumption that the prescriber likely has more information about a patient’s unique situation.46 The number of times the medication had previously been filled was a significant vignette factor with pharmacists more interested in asking side effect and effectiveness questions early on (Table 3). This supports the qualitative finding that for some pharmacists, once the patient has filled the prescription multiple times, they have an “If it aint broke, don’t fix it” mentality.20 This may not be ideal as adherence rates decrease over time47 and side effects or safety issues can present beyond the initiation period of a medication, especially as medications are added or changed and as patients increase in age. This tendency is consistent with the findings that pharmacists are more likely to counsel on new prescriptions than refills14 and focus on information giving rather than question asking.16 Creating mechanisms to remind, incentivize, and support pharmacists for long term follow up is an area for future research. Past interaction and relationship factors The pharmacist-patient relationship was not well modeled in the current vignette approach, despite its prominence in the qualitative study.20 A more vivid description than a patient “whom you know by name” may be necessary to test the importance of past interaction in question asking likelihood using this method. Other research methods may be better suited for

648

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650

exploring the role of patient-pharmacist relationships on processes and outcomes. It is important to continue to incorporate the pharmacistpatient relationships because of their role in quality care.20,48,49 Limitations The factorial vignettes were not intended to test behaviors, rather, they were used to assess judgments of behavioral intention, which, under certain conditions, have been associated with behavior.50,51 Also, the orthogonal nature of the vignettes forces some variables to be associated more or less frequently than in reality. In addition, the survey did not assess the extent to which respondents found the vignettes to be realistic. The likelihood numbers supplied by pharmacists may have been influenced by norms or ideals. Regardless, this study provides a starting point for investigating medication monitoring decision making, and attitudes in community pharmacy. Some respondents put the same likelihood for all three question types, suggesting satisficing, which may have made parameters more difficult to estimate. Asking more varied questions may help engage more respondents. Also, including more vignettes per respondent would help with the power to detect significant variation at the participant level. Lastly, pharmacists were only sampled from one Midwestern state which affects generalizability and non-participants may have different characteristics and may give different responses. Practice implications “Days late” is potentially one of the most accessible pieces of medication monitoring information available to community pharmacists. Pharmacists, however, may not be using this information to initiate discussions about nonadherence in practice.20 Technicians or computer alerts could be used to cue pharmacists when patients present to pick up late prescriptions given the asynchronous sequence of when prescriptions are requested, evaluated, filled, and picked up. Question guides can be used to promote pharmacists engaging in comprehensive discussions about patients’ reasons for non-adherence when nonadherence is identified.52 Pharmacies, however, are busy and this affects pharmacists’ tendencies to monitor medications and this must be taken into consideration when designing nonadherence interventions.

The medication-specific differences identified in this study suggest that different medications may benefit from different monitoring approaches, for example, asking about depression symptoms when monitoring antidepressant medications.53,54 Certain medication classes may need to be actively targeted because pharmacists are not allocating their attention proportional to the potential positive impact. More research is needed in this area. A one-size-fits-all approach may not be ideal for addressing the large variety of prescriptions dispensed at pharmacies.

Conclusions The factorial survey method was a useful approach for assessing pharmacists’ mental model for deciding when to ask monitoring questions at the time of refill. Pharmacists appear to incorporate specifics of the medication and the number of days late (or early for controlled substances) when deciding when to engage in monitoring. The pharmacists’ attitudes about their monitoring role also was important. Environmental barriers such as the pharmacy being short staffed or there being a queue of patients waiting were deterrents to self-reported monitoring likelihood. Supports are needed to bolster pharmacist monitoring given the varied values they place on different medications and the environmental barriers to monitoring. Acknowledgments To Drs. Bernard Sorofman, Erika Ernst, Pamela Wesley, Amber Goedken, and Christopher Lyons for study suggestions and feedback.

References 1. Sarkar U, Lo´pez A, Maselli JH, Gonzales R. Adverse drug events in US adult ambulatory medical care. Health Serv Res 2011;46:1517–1533. 2. Budnitz DS, Lovegrove MC, Shehab N, Richards CL. Emergency hospitalizations for adverse drug events in older Americans. N Engl J Med 2011;365:2002–2012. 3. Steinman MA, Handler SM, Gurwitz JH, Schiff GD, Covinsky KE. Beyond the prescription: medication monitoring and adverse drug events in older adults. J Am Geriatr Soc 2011;59:1520–1530. 4. Forster AJ, Clark HD, Menard A, et al. Adverse events among medical patients after discharge from hospital. Can Med Assoc J 2004;170:345–349.

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650 5. Forster AJ, Murff HJ, Peterson JF, Gandhi TK, Bates DW. Adverse drug events occurring following hospital discharge. J Gen Intern Med 2005;20: 317–323. 6. Gandhi TK, Weingart SN, Borus J, et al. Adverse drug events in ambulatory care. N Engl J Med 2003;348:1556–1564. 7. Budnitz DS, Pollock DA, Weidenbach KN, Mendelsohn AB, Schroeder TJ, Annest JL. National surveillance of emergency department visits for outpatient adverse drug events. J Am Med Assoc 2006;296:1858–1866. 8. Budnitz DS, Shehab N, Kegler SR, Richards CL. Medication use leading to emergency department visits for adverse drug events in older adults. Ann Intern Med 2007;147:755–765. 9. NEHI. Thinking Outside the Pillbox: A System-wide Approach to Improving Patient Medication Adherence for Chronic Disease; 2009. 10. Qiuping G, Dillon CF, Burt VL. Prescription Drug Use Continues to Increase: U.S. Prescription Drug Data for 2007–2008. Hyattsville, MD: National Center for Health Statistics; 2010. 11. Murphy P, Cocohoba J, Tang A, Pietrandoni G, Hou J, Guglielmo BJ. Impact of HIV-specialized pharmacies on adherence and persistence with antiretroviral therapy. AIDS Patient Care STDs 2012; 26:526–531. 12. Tsuyuki RT, Johnson JA, Teo KK, et al. A randomized trial of the effect of community pharmacist intervention on cholesterol risk management: the Study of Cardiovascular Risk Intervention by Pharmacists (SCRIP). Arch Intern Med 2002;162:1149– 1155. 13. Kreling DH, Doucette WR, Chang EH, Gaither CA, Mott DA, Schommer JC. Practice characteristics of bachelor of science and doctor of pharmacy degreed pharmacists based on the 2009 National Workforce survey. Am J Pharm Educ 2010;74:1–8. 14. Puspitasari HP, Aslani P, Krass I. A review of counseling practices on prescription medicines in community pharmacies. Res Social Adm Pharm 2009; 5:197–210. 15. Sleath B. Pharmacist question-asking in New Mexico community pharmacies. Am J Pharm Educ 1995;59:374–376. 16. de Oliveira DR, Shoemaker SJ. Achieving patient centeredness in pharmacy practice: openness and the pharmacist’s natural attitude. J Am Pharm Assoc 2006;46:56–66. 17. Murad MS, Chatterley T, Guirguis LM. A metanarrative review of recorded patient–pharmacist interactions: exploring biomedical or patient-centered communication? Res Social Adm Pharm 2013;10:1–20. 18. Ha¨mmerlein A, Griese N, Schulz M. Survey of drug-related problems identified by community pharmacies. Ann Pharmacother 2007;41:1825– 1832.

649

19. Svarstad BL, Bultman DC, Mount JK. Patient counseling provided in community pharmacies: effects of state regulation, pharmacist age, and busyness. J Am Pharm Assoc 2004;44:22–29. 20. Witry MJ, Doucette WR. Community pharmacists, medication monitoring, and the routine nature of refills: a qualitative study. J Am Pharm Assoc 2014;54:594–603. 21. Svarstad BL, Bultman DC. The patient: behavioral determinants. In: Gennaro AE, ed. Remington’s the science and practice of pharmacy, vol. 20. Baltimore, MD: Lippincott, Williams, and Wilkins; 2000. 22. Stevenson FA, Barry CA, Britten N, Barber N, Bradley CP. Doctor-patient communication about drugs: the evidence for shared decision making. Soc Sci Med 2000;50:829–840. 23. Creswell JW, Plano-Clark VL. Designing and Conducting Mixed Methods Research. Thousand Oaks, CA: Sage; 2007. 24. Rossi PH, Anderson AB. Measuring Social Judgements: The Factorial Survey Approach. Beverly Hills, CA: Sage; 1982. 25. Louviere JJ, Flynn TN, Carson RT. Discrete choice experiments are not conjoint analysis. J Choice Model 2010;3:57–72. 26. Watson MC, Norris P, Granas AG. A systematic review of the use of simulated patients and pharmacy practice research. Int J Pharm Pract 2006;14:83–93. 27. Roter D, Lipkin M, Korsgaard A. Sex differences in patients’ and physicians’ communication during primary care medical visits. Med Care 1991;29: 1083–1093. 28. Phokeo V, Sproule B, Raman-Wilms L. Community pharmacists’ attitudes toward and professional interactions with users of psychiatric medication. Psychiatr Serv 2004;55:1434–1436. 29. Vainio KK, Airaksinen MS, Hyykky TT, Enlund KH. Effect of therapeutic class on counseling in community pharmacies. Ann Pharmacother 2002;36:781–786. 30. Dillman DA, Smyth JD, Christian LM. Internet, Mail, and Mixed-mode Surveys: The Tailored Design Methodvol. 3. New York: Wiley; 2008. 31. Witry MJ. Community pharmacist medication monitoring attitudes and decision making. The University of Iowa. Proquest, UMI Dissertations Publishing; 2013. 3608514. 32. Hox JJ, Kreft IGG, Hermkens PLJ. The analysis of factorial surveys. Sociol Methods Res 1991;19:493–510. 33. Jasso G. Factorial survey methods for studying beliefs and judgments. Sociol Method Res 2006;34: 334–423. 34. Snijders TA. Multilevel Analysis. New York: Springer; 2011. 35. Schommer JC, Wiederholt JB. Pharmacists’ views of patient counseling. Am Pharm 1994;7:46–53. 36. Hall JA, Irish JT, Roter DL, Ehrlich CM, Miller LH. Gender in medical encounters: an analysis of physician and patient communication in a

650

37.

38.

39.

40.

41.

42.

43. 44.

45.

Witry & Doucette / Research in Social and Administrative Pharmacy 11 (2015) 639–650 primary care setting. Health Psychol 1994;13: 384–392. Cocohoba J, Comfort M, Kianfar H, Johnson MO. A qualitative study examining HIV antiretroviral adherence counseling and support in community pharmacies. J Manag Care Pharm 2013;19:454. Svarstad BL, Chewning BA, Sleath BL, Claesson C. The brief medication questionnaire: a tool for screening patient adherence and barriers to adherence. Patient Educ Couns 1999;37:113–124. Odukoya OK, Chui MA, Pu J. Factors influencing quality of patient interaction at community pharmacy drive-through and walk-in counselling areas. Int J Pharm Pract 2014;22:246–256. Shah BK. Pharmacists’ Work Environment and Their Practice Behaviors. University of WisconsinMadison; 2006. Schommer JC, Wiederholt JB. Pharmacists’ perceptions of patients’ needs for counseling. Am J Hosp Pharm 1994;51:478–485. Bitner MJ. Evaluating service encounters: the effects of physical surroundings and employee responses. J Mark 1990;54:69–82. Nau DP. Measuring pharmacy quality. J Am Pharm Assoc 2009;49:154–163. Chhobanian AV, Bakris GL, Black HR, et al. The seventh report of the Joint National Committee on prevention, detection, evaluation, and treatment of high blood pressure. J Am Med Assoc 2003;289: 2560–2571. DiMatteo MR, Lepper HS, Croghan TW. Depression is a risk factor for noncomploiance with medical treatment. Arch Intern Med 2000;160:2101–2107.

46. Losey CS, Hoehns JD, Schlobohm C, Witry MJ. Investigation of simulated pharmacist decision making involving prescriptions with a high probability of causing patient harm. J Pharm Tech 2014; 30:207–215. 47. Caro JJ, Salas M, Speckman JL, Raggio G, Jackson JD. Persistance with treatment for hypertension in actual practice. Can Med Assoc J 1999; 160:31–37. 48. Worley MM. Testing pharmacist-patient relationship quality model among older persons with diabetes. Res Social Adm Pharm 2006;2:1–21. 49. Doucette WR, McDonough RP. Beyond the 4Ps: using relationship marketing to build value and demand for pharmacy services. J Am Pharm Assoc 2002;42:183–194. 50. Ajzen I. The theory of planned behavior. Organ Behav Hum Decis Process 1991;50:179–211. 51. Kim M, Hunter JE. Relationships among attitudes, behavioral intentions, and behavior: a metaanalysis of past research (Pt. 2). Commun Res 1993;20:331–364. 52. Witry MJ, Doucette WR, Zhang Y, Farris KB. Multiple adherence tool evaluation study (MATES). J Manag Care Pharm 2014;20:734–740. 53. Bultman DC, Svarstad BL. Effects of pharmacist monitoring on patient satisfaction with antidepressant medication therapy. J Am Phar Assoc 2005; 42:36–43. 54. Rickles NM, Svarstad BL, Statz-Paynter JL, Taylor LV, Kobak KA. Pharmacist telemonitoring of antidepressant use: effects on pharmacistpatient collaboration. J Am Pharm Assoc 2005; 45:344–353.

Factors influencing community pharmacists' likelihood to ask medication monitoring questions: A factorial survey.

Community pharmacists are well positioned to identify and resolve medication related problems associated with chronic medication use during prescripti...
516KB Sizes 0 Downloads 6 Views