Utilization BY .\I)
Review
LARRY l)A\Il)
KIIISTEIN. \l.l).. \I. I)IIESSLER.
of Treatment BRIGITTE \1.I).
for Suicide PRI
SOFF.
\1.P.E1..
The authors describe an effort to develop criteria for utilization revie w of treat ment for suicide attempters. Explicit criteria proposed by a panel ofexperts as essential determinants for hospitalization of these patients were compared with actual clinical practice. It wasfound that according to the experts’ criteria (which were operationalized into rating assessments), over half ofihe outpatient sample should have been hospitalized. After multiple regression analysis was carried out on the criteria, however, four predictors showed that onlt’ 20 percent ofthe outpatients should have been hospitalized. The authors discuss the issues thesefindings raise about the criteria ofthe experts, their utilitt’for research, their validity, and their implications for utilization review.
UTILIZATION REVIEW, which is now a standard part of medical practice, has only recently been introduced into psychiatric cane (1-4). The difficulties in monitoring the quality of psychiatric care are major. There is lack of agreement about what type of information is relevant, how the information should be collected and interpreted, and what the appropriate outcome measures are. In 1969 the psychiatric utilization review and evaluation project (PURE) was undertaken by a group cornposed of members of the departments of psychiatry, epidemiology, public health, and sociology of Yale University and members of the psychiatric staff from the Connecticut Mental Health Center. This group began the task ofdeveloping a basic utilization review program for patient cane evaluation in a community mental health center. Various ways of approaching the problem of creating a mechanism for the continuous monitoring of the propriety, effectiveness, and quality of patient care services
Dr. Kirstein is a Postdoctoral Fellow in Psychiatry, Yale University School of Medicine, 34 Park St., New Haven, Conn. 06508, and the Connecticut Mental Health Center, New Haven, Conn. Ms. Prusoff and Dr. Weissman are Assistant Professors, Yale University School of Medicine, and Directors, Depression Research Unit, Connecticut Mental Health Center. Dr. Dressier is Chief of Psychiatry, New Britain General Hospital, New Britain, Conn., and Assistant Professor of Psychiatry, University of Connecticut School of Medicine, Hartford, Conn. This study was supported in part by contract National Institute of Mental Health.
22
A
mJ
Psychiatry 132:1,January
MSM-42-69-60
1975
from
the
Attempters \IYRNA
\\ EISSMA\.
Pill).,
were suggested. They included the following: 1) development of screening devices to select cases for review; 2) development of adequate guidelines (or explicit criteria) by panels of experts for the evaluation of care; 3) development of more adequate instruments for data collection; 4) assessment of factors involved in the actual process of evaluation of care by the reviewing physician; and 5) use of epidemiological surveys and follow-up studies to moniton patient care beyond the period of the patient’s institutional stay. As part of the project various panels were formed, one of which was organized to develop criteria for evaluation of cane given patients with suicidal behavior. This panel used the criterion-oriented approach described in detail by Tischlen and Riedel (5). Based on a review of patients’ charts and on their own clinical experience and knowledge of the literature, the panel developed implicit norms that were then translated into explicit criteria for the hospitalization of patients exhibiting suicidal behavior. They developed nine criteria for which suicide attempters should be hospitalized and conducted a prospective study to compare these criteria with actual clinical practice. This paper has several purposes. We intend to describe the nine criteria and how they were translated into operationalized assessments that could then be studied, to test the validity of the criterion-oriented approach by comparing the explicit criteria for psychiatric hospitalization with actual hospitalizations, and to explore alternate ways of refining the explicit criteria so that they might be more suitable for utilization review.
METHOD
Treatment
Setting
and Criteria
for
Selection
The prospective study took place in New Haven, Conn., at the Yale-New Haven Hospital and the Connecticut Mental Health Center. The main source of patients was the emergency room of the hospital, which at the time of the study provided the main resource for acute treatment for suicide attempters in the New Haven area. The staff were linked with the emergency services of the adjacent mental health center. During the day, crisis patients might be seen at either resource; at night they were seen only in the emergency room. The study included all patients aged 16 or oven who
KIRSTEIN,
came or were brought to the emergency services with a history of a suicide attempt or a suicidal gesture within the past three days. A suicide attempt was defined as any intentionally self-inflicted injury (including self-injuries by ingestion) unless there was good evidence both in the circumstances and in the patient’s statements that there was not even partial or ambiguous self-destructive intent. Lack ofintent was not assumed merely on the basis of the patient’s denial, absence of serious risk to life, or added manipulative elements. It was accepted in cases such as attempted abortion by pill ingestion and cases of cornpulsive minor self-mutilation. Cases ofaccidental self-injury in which unconscious self-destructive intent was suspected were excluded in our definition of a suicide attempt because the injury was not intentional. Included in our definition were suicidal gestures in which initiating moves of a potentially self-destructive nature had been carried out even if there was no risk of significant physical injury, e.g., climbing on a bridge, toying with a gun, announcing intent to ingest pills but actually taking no more than one or two. Some kind of act was required; patients only complaining of suicidal feelings, no matter how intensely, were not included. Screening
Procedures
and Sampling
Screening took place over a period of 25 weeks, from November 1970 to May 1971. During this time psychiatnc residents were instructed to identify and record data on all eligible subjects appearing at any time ofthe day on night on all days of the week. As a check on this procedune a research assistant examined emergency room records and other admission records daily to identify missed subjects. The appropriate resident was immediately contacted and the relevant information obtained in the event of missed cases. During the 25 weeks, 274 subjects were identified and at least some data were obtained on each of them. The average inflow was I 1 cases a week. Twenty-five patients hospitalized for medical sequelae of suicide attempts were excluded from the study because the disposition choice was a medical unit, and one patient was excluded because of insufficient data. This left a sample of 248 suicide attempters. Characteristics of the sample and the methodology have been detailed elsewhere (6). Data
Collection
Data were collected primarily by the psychiatric resident who first evaluated the patient clinically. After pilot studies had been carried out a precoded data-recording form was devised.’ This form contains 101 items, including sociodemognaphic status, previous history, details of the attempt,judged and acknowledged motivations, mental status, social circumstances, what treatment was instituted, and residents’ feelings toward the patient. Definitions of items and anchor points for each item’s description rating were given in as much detail as possible. Items were adapted from a variety of sources, particularly the Los Angeles Suicide Prevention Center As‘This
form
is available
from
the
authors
on request.
PRUSOFF,
WEISS
MAN,
AND
DRESSLER
sessment of Suicidal Potentiality (7), and, for mental status, a modified version of the Hamilton Rating Scale for Depression (8). The form was completed by the psychiatric resident during or immediately after the initial interview with the patient. Almost all suicidal patients were seen by a psychiatnist when they first came to the emergency unit. For patients who were too physically ill or drowsy at this time, information was obtained after admission. Vanables that referred only to the time of admission, such as mental status, were omitted. Thirty-two of the original 274 patients ( 1 1.6 percent) were discharged or died before information was obtained. For these patients, only basic sociodemographic data and treatment disposition were recorded. All of the data collected were recorded on precoded forms suitable for computer analysis.
RESULTS
Table I shows the criteria patients with suicidal behavior panel as well as the translation ationalized rating assessments tenia. These were administered interview. Most of the criteria but there was some redundancy. Disposition
for hospitalization of developed by the expert of these criteria into openand operationalized cnias part of the resident’s could be operationalized,
Groups
The 18 operationalized criteria were rated on different scales, most of which were 4-point scales. Each scale was divided into two groups (e.g., high or not high, present or absent) according to the clinicians’ judgment of the scale’s reflection of the clinical situation. This dichotomization was done by a psychiatrist (L.K.) before the data were analyzed and was independent of the prospective study. A cross-tabulation of each of the 18 items was made with the three possible treatment dispositions. Table 2 compares the operationalized explicit criteria with the treatment actually recommended. The items are ordered by their power to discriminate among the three groups. Suicidal risk was the single most powerful item determining whether the suicidal patient was to receive a necommendation of hospitalization. It was found that 80.8 percent of the 121 patients who were hospitalized were considered high suicidal risks while only 25.7 percent of the 127 patients who were not hospitalized were considered suicidal risks. Eleven of the 18 items showed a significant prediction in the direction that the expert panel hypothesized. However, item 4b, “change from usual state,” was a significant predictor in the opposite direction. A person having this symptom was more likely to have had an actual disposition of outpatient treatment than to have been hospitalized, although the experts judged that the presence of this behavior requires hospitalization. The expert panel had agreed that the presence of any one of the criteria necessitated hospitalization. It is evident from the data shown in table 2 that all of the suicidal
Am J Psychiatry
132.’!, January
1975
23
UTILIZATION
TABLE The
REVIEW
FOR
SUICIDE
ATTEMPTERS
I Explicit
Experts’
Experts’ 1 A clear
Explicit lethal
Criteria.
Their
Operationali:ation
Criterion suicide
plan
What
were
Id
Using ofthe Taking risk of What What attempt?
all available data from the interview. what is assessment patient’s intent to kill himself’? into account all ofthe evidence, what was the potential the attempt to the patient’s life? were the medical etTects of the attempt? is the patient’s acknowledged overt intent in making the
was the plan for the attempt has recent suicidal behavior the
social
2a
Hos many attempts including this present
2h
How
many
life,
including
circumstances
attempt?
5
6 The presence of high-risk social circumstances, such as social isolation 7 A current, clear suicide attempt 8 Expression of strong suicidal thoughts with intent and without seeing any other way out 9 Expectation of change in behavior of significant others due to suicidal behavior not met: change cannot be accomplished appropriately were generally
4-point
suicide
this
gestures
present
What
6 7a
is assessment
in the
of the
near
the
in his entire
patient
Criterion
Specific plan Intensity of attempt Help sought Serious
intent
Serious
risk
Major Overt
medical intent
eflects
made
in his
life.
behavior
show
the
Previous
attempts
Previous
gestures
entire
episode?
following Bizarre motivation Depressive delusions Paranoid delusions Thought disorder
risk
of suicidal
future?
behavior
by
Speed
of onset
Change
from
Plans
usual
state
delayed
this
Suicidal
What are the patient’s living arrangements? through 7g: same as Ia through Ig
Living Sameas
risk
arrangements lathrough
Ig
None
None
scales
with definite
anchor
points.
patients in this study had at least one of the symptoms and that according to the experts’ criteria all 248 patients should have been hospitalized. Thus the 127 patients (51 percent) who were not hospitalized would have, by the experts’ criteria, been subject to utilization review. Regression
has
made
How rapid was the onset and development of the present suicidal preoccupation? How much do recent suicidal feelings and preoccupation represent a change from the patient’s usual state? How long had the patient delayed betseen deciding to take his life and making the attempt?
patient
Approach
In an effort to determine which combinations of items were most important to determine disposition in actual clinical practice, all 18 items were entered into a stepwise multiple regression analysis. Disposition was transformed into a dichotomous variable: the inpatients were given a score of 1 and noninpatients a score of 2.
Am
and its execution? been?
ofthe
has the patient episode?
what extent does the suicidal niotivations: 3a Bizarre motivation? 3b Depressive delusions? 3c Paranoid delusions? 3d Thought disorder?
4a
An expectation of hospitalization that cannot be changed during interview
24
Operationalized
Ic
4c
Multiple
Assessments*
specific intense
4b
scales
Criteria
Hos How
To
4 A recent progression in seriousness of thoughts from suicidal thoughts to gestures
rating
Rating
and Operationali:ed
of medicaIl
3 The presence of suicidal thoughts. gestures, or attempts in association with delusions or psychosis
*The
of
Assessments.
Ia lb
II lg
5
Rating
Operationalization
Ic
2 A recent history serious attempts
into
J Psychiatry 132:!,January
1975
Four items were determined to be significant by the multiple regression analysis: suicidal risk, thought disorder, serious intent, and major medical effects. These items accounted for 38 percent ofthe variance in predicting disposition. It must be noted that the suicidal risk item accounted for 31 percent of the variance; the value of the three additional items was therefore small. Using all four items, 51 of 248 patients (20.5 percent) were considered misclassified: 15 hospitalized patients who were predicted not to require hospitalization by the criteria and 36 nonhospitalized patients who were predicted to require hospitalization. Thus, using these four items, the chants of 51 patients would be open to utilization review.
KIRSTEIN,
TABLE
PRUSOFF,
WEISSMAN,
AND
DRESSLER
2
(‘omparison
of Operationali:ed
E.vplicit
Criteria
for Hospitalization
and A ctual
Actual No Further Treatment Criterion
Number
Suicidal risk High Low Serious intent Serious Not serious Overt intent Overt Notovert Previous attempts One or more None Thought disorder Present Absent Serious risk High Loss Intensity of attempt High Low Bizarre motivation Present Absent Specific plan Present Absent Major medical effects Yes No Plan delayed Yes No Living arrangements Bad Good Depressive delusions Present Absent Paranoid delusions Present Absent Previous gestures One or more None Help sought No Yes Speed of onset Rapid Not rapid Change from usual state Yes No Any of the above Present Absent
Disposition
(N
248i
Disposition Outpatient Treatment
Percent
Number
ChiSquare
Hospitalization Percent
Number
Percent
Value*
8 13
38.1 61.9
27 78
25.7 74.3
97 23
80.8 19.2
7#{216}7**
3 IS
14.3 85.7
3 100
2.9 97.1
34 86
28.3 71.7
26.2**
6 16
27.3 72.7
43 59
42.2 57.8
79 34
69.9 30.1
23.6**
I 9
5.0 95.0
17 86
l6.5 83.5
47 65
42.0 58.0
23.0**
I 20
4.8 95.2
2 102
1.9 98.1
25 93
21.2 78.8
2 I .2
4 l8
18.2 81.8
6 98
5.8 94.2
35 86
28.9 71.1
20.l**
2 19
9.5 90.5
7 97
6.7 93.3
33 86
27.7 72.3
l8.l**
I 20
4.8 95.2
5 97
4.9 95.1
25 91
21.6 78.4
I4.7**
3 IS
14.3 85.7
5 99
4.8 95.2
26 90
22.4 77.6
14.0**
I 21
4.5 95.5
0 105
100.0
10 111
8.3 9l.7
9.l***
6 14
30.0 70.0
28 76
26.9 73.1
50 60
45.5 54.5
8.3***
3 19
13.6 86.4
II 94
10.5 89.5
25 94
21.0 79.0
4.7
I
4.8 95.2
3 101
2.9 97.1
9 103
8.0 92.0
2.8
I 20
4.8 95.2
4 100
3.8 96.2
8 103
7.2 92.8
1.2
4 17
19.0 81.0
21 82
20.4 79.6
28 83
25.2 74.8
0.9
I
4.8 95.2
3 100
2.9 97.1
4 115
3.4 96.6
0.2
20 0
100.0
-
96 8
92.3 7.7
98 14
87.5 12.5
3.7
19 2
90.5 9.5
83 21
79.8 20.2
65 49
57.0 43.0
18.1
22 0
100.0
105 0
100.0
121 0
l00.0
20
20
-
-
**
*df ***p