| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Sandra F. Simmons, Jewish Home for the Aging/Borun Center for Gerontological Research, UCLA, 7150 Tampa Ave., Reseda, CA 91335. E-mail: ssimmons{at}ucla.edu
 |
Abstract |
|---|
 TOP Abstract Design and MethodsResultsDiscussionReferences |
|---|
Design and Methods. A cross-sectional study with 396 long-term residents in 14 skilled nursing facilities was conducted: 10 homes in the lower (25th percentile: low prevalence 0–2%) quartile and 4 homes in the upper (75th percentile: high prevalence 12–14%) quartile on the MDS depression quality indicator. Ten care processes related to depression were defined and operationalized into clinical indicators. Measurement of nursing home staff implementation of each care process and the assessment of depressive symptoms were conducted by trained research staff during 3 consecutive 12-hr days (7 a.m. to 7 p.m.), which included resident interviews (Geriatric Depression Scale), direct observations, and medical record review using standardized protocols.
Results. The prevalence of depressive symptoms according to independent assessments was significantly higher than prevalence based on the MDS quality indicator and comparable between homes reporting low versus high rates of depression (46% and 41%, respectively). Documentation of depressive symptoms was significantly more common in homes reporting a high prevalence rate; however, documentation of symptoms on the MDS did not result in better treatment or management of depression according to any care-process measure. Psychosocial prevention and intervention efforts, such as resident participation in organized social group activities, were not widely used within either group of homes.
Implications. The MDS depression quality indicator underestimates the prevalence of depressive symptoms in all homes but, in particular, among those reporting low or nonexistent rates. The indicator may be more reflective of measurement processes related to detection of symptoms than of prevention, intervention, or management of depression outcomes. A depression quality indicator should not be eliminated from MDS reports because of the importance and prevalence of the condition. However, efforts to improve nursing home staff detection of depressive symptoms should be initiated prior to the use of any MDS-based depression indicator for improvement purposes. Homes that report a low prevalence of depression according to the nationally publicized MDS quality indicator should not be regarded as providing better care.
Key Words: Nursing homes • Quality of care • Mental health
One assumption is that differences between facilities on the MDS-derived quality indicators reflect differences in care-process implementation by nursing home staff (e.g., prevention, treatment), which in turn explain differences in clinical outcomes among residents (e.g., prevalence of depression). If this assumption is valid, then the advantage of providing quality-indicator information to nursing home staff is to provide feedback to staff about care-process implementation to facilitate the identification of areas in need of improvement. Consumers and survey staff could use quality-indicator information to identify nursing homes that might be providing different levels of care quality.
Depression is a common problem that is often undetected and undertreated among nursing home residents (Abrams, Teresi, & Butin, 1992; Parmelee, Katz, & Lawton, 1989; Rovner, German, Brant, Clark, Burton, & Folstein, 1991; Schnelle, Wood, Schnelle, & Simmons, 2001). Prevalence rates of major depression in the long-term-care setting have been reported to range from 10% to 22%, whereas an additional 15–50% of residents may show significant depressive symptoms (Abrams et al., 1992; Parmelee et al., 1989; Schnelle et al., 2001). The MDS user's manual states, within the mood assessment guidelines, "about 15% of nursing home residents will have a major depression; about 30% will exhibit noticeable symptomatic signs of a mood state problem" (HCFA, 1999, p. 88). The detection of depressive symptoms among nursing home residents is important, even if the symptoms do not meet criteria for major depression, because the presence of symptoms should prompt nursing home staff to conduct appropriate follow-up assessments and to identify residents in need of intervention to prevent worsening of symptoms (HCFA, 1999). Major depression is associated with increased functional dependence, health-care utilization, and mortality (Parmelee, Katz, & Lawton, 1992; Rovner et al., 1991).
The accuracy of the MDS depression quality indicator was evaluated in a recent study that compared two nursing homes (Schnelle et al., 2001). One nursing home had a low MDS-based prevalence rate of depression (1%), which placed the home in the lower 25th percentile (good resident outcome); the other nursing home had a high prevalence rate (12%), which placed the home in the upper 75th percentile (poor resident outcome). Independent evaluations by research staff of the prevalence of probable depression using a standardized interview measure (i.e., Geriatric Depression Scale [GDS]), however, showed significantly higher prevalence rates of probable depression in both the low and the high home (49% and 55%, respectively), compared with MDS data, and no difference between homes (Schnelle et al., 2001). The results of this preliminary study indicate that the prevalence of depression as reported on the MDS may reflect differences between homes related to the detection and documentation of depressive symptoms by indigenous nursing home staff (Schnelle et al., 2001). Specific care processes related to the treatment and/or management of depression were not evaluated in this study.
One potential explanation for the underdetection of depressive symptoms by indigenous nursing home staff is that the MDS items used to score the quality indicator are based primarily on nursing home staff observation of depressive behaviors instead of residents' responses to standardized screening questions. The MDS manual instructs nursing home staff to "conduct observations throughout the day and initiate a conversation" with the resident. However, staff are not instructed to use specific questions, nor are they required to use a standardized mood assessment (HCFA, 1999). There is evidence that trained research staff are able to reliably complete MDS items related to depression (Burrows, Morris, Simon, Hirdes, & Phillips, 2000). However, other studies have shown that symptoms of depression as documented on the MDS by indigenous nursing home staff did not correlate well with independent assessments of depressive symptoms using interview scales (Frederiksen, Tariot, & De Jonghe, 1996; Schnelle et al., 2001). The results of a separate study showed that the detection of depressive symptoms (e.g., sad affect, crying, negative self-statements) based on observation was poor among both licensed and unlicensed nursing home staff even when staff used a standardized behavioral observation scale (Wood et al., 2000). The use of direct questions to identify the presence of depressive symptoms among nursing home residents is further supported by the results of several recent studies, which have shown that a substantial proportion of nursing home residents with mild to moderate cognitive impairment are capable of reliably answering interview questions about their mood state using a standardized interview such as the GDS (Gerety et al., 1994; Schnelle et al., 2001; Sheikh & Yesavage, 1986).
The purpose of the current study was to replicate and extend the results of a previous study (Schnelle et al., 2001). Specifically, two research questions were addressed in this study: (a) Do the results of independent assessments of depressive symptoms by research staff show differences in prevalence of depression for participants in groups of homes that scored in the lower (25th percentile: low prevalence) or upper (75th percentile: high prevalence) quartile on the MDS-based prevalence of depression quality indicator? (b) Are there differences in care-process implementation related to assessment and treatment of depression between homes that scored in the extreme quartiles on the MDS depression quality indicator?
Care processes related to depression were defined and operationalized into clinical indicators for the purpose of this study based on the expert consensus results of the Assessing the Care of the Vulnerable Elderly (ACOVE) Project (Shekelle, MacLean, Morton, & Wenger, 2001; Wenger & Shekelle, 2001). Standardized medical record review, direct observation, and resident-interview protocols were developed to measure nursing home staff implementation of the defined care processes related to depression, and specific scoring rules were developed that allowed statements to be made about the quality of depression care within facilities.
|
Design and Methods |
|---|
|
TOPAbstractDesign and MethodsResultsDiscussionReferences |
|---|
Participants were recruited over a 2-week period at each nursing home site. The only residents excluded from the study were those in transitional care (i.e., Medicare-covered stays). Written consent was obtained from either the resident, if capable of self-consent, or his or her "responsible party" designated in the medical record. The "responsible party," typically a family member, had to be contacted for the majority (68%) of residents to obtain consent. Four hundred twenty-six residents provided consent (157 residents capable of self-consent + 269 respective family members), with an average consent rate of 35% per home, and 396 residents completed the study.
Measures
Demographic information, including age, length of residency, gender, ethnicity, and MediCal payment status, was retrieved from each participant's medical record. MDS information related to mood, cognitive functioning, hearing impairment, and vision impairment was retrieved from the most recently completed MDS assessment at the time of the study (HCFA, 1999). Medical record reviews related to depression were conducted according to a standardized protocol for a subsample of 269 participants by a trained research physician and/or geriatric nurse practitioner, both of whom were blind to the placement of the home in either the upper or the lower quartile. Trained research staff, also blind to the placement of the homes, conducted direct observations and resident interviews during an assessment period of 3 consecutive weekdays (12 hours/day, 7 a.m. to 7 p.m.) within each nursing home for all participants.
The term "quality indicator" is used in this study to refer to the MDS-derived prevalence of depression measure as an indicator of overall nursing home-care quality related to depression. The term "clinical indicator" refers to the defined care processes (derived from the ACOVE Project) that were used in this study to measure the quality of assessment, prevention, and management related to depression.
MDS Information
The depression quality indicator is scored based on 14 MDS items (Zimmerman et al., 1995). The MDS depression quality indicator was calculated for the consented samples within each nursing home based on the most recent MDS data at the time of the study. MDS information was also retrieved for each of the 16 items relevant to the assessment of disturbed mood (MDS Section E, Mood and Behavior Patterns, items 1a–p). Nursing home staff documentation of the presence of any 1 of the 16 mood items for a resident necessitates a follow-up Resident Assessment Protocol (MDS-RAP) for mood disturbance (HCFA, 1999).
Resident Interview
Research staff conducted independent assessments of depressive symptoms using the GDS short form (GDS-short). The 15-item short form of the GDS has been specifically recommended by one practice guideline as a screen for probable depression in nursing home residents (American Medical Directors Association, 1996). The GDS-short has a total score range from 0 to 15 in which a score of >5 is indicative of probable depression (Sheikh & Yesavage, 1986). The sensitivity and specificity of the GDS-short in detecting probable depression have been shown to be reduced among individuals with cognitive impairment, in particular among those with a Mini-Mental State Examination (MMSE) total score of <15 (McGivney, Mulvihill, & Taylor, 1994). In consideration of the possible influence of dementia on the reliability of the GDS-short scores, participants with an MDS recall score of 
2, which has been shown to be equivalent to an MMSE total score of 17 (Simmons & Schnelle, 2001), were interviewed with the GDS-short measure in this study.
In addition, an attempt was made to conduct a second GDS-short interview for a convenience sample of 37 participants who completed the first interview to evaluate the stability of resident-interview responses. The second GDS-short interview was attempted on the day following the initial GDS-short assessment (i.e., 2 consecutive assessment days). Of the repeat interview attempts with 37 participants, 34 participants provided complete information for the second interview, and classification as "probable depression" (i.e., total score of >5) showed good stability among this subsample (
=.68, p <.001). Interrater reliability data were collected for a different subsample during the GDS interview and showed excellent agreement (n = 43, value range for GDS yes/no items =.83–1.0, p <.01).
Direct Observations
Direct observations of depressive symptoms were conducted during the GDS-short interview to document the presence of behavioral symptoms of depression. The rationale for including the direct observations of behavioral symptoms was that the direct interview questions about mood might elicit behavioral symptoms of depression (e.g., crying, tearfulness). The behavioral symptoms of disturbed mood were directly derived from the following MDS items (Section E, Mood and Behavior Patterns, 1. Indicators of depression, anxiety, sad mood, items a–j, l–n): (a) negative statements, (b) repetitive questions, (c) repetitive verbalizations, (d) persistent anger with self or others, (e) self-deprecation, (f) expressions of what appear to be unrealistic fears, (g) recurrent statements that something terrible is about to happen, (h) repetitive health complaints, (i) repetitive anxious complaints and concerns, (j) unpleasant mood in morning, (l) sad, pained, worried facial expression, (m) crying, tearfulness, and (n) repetitive physical movements. Item k is related to insomnia for which there was a direct question on the interview. Items o and p are related to social engagement, which was assessed using a different observational protocol outside of the interview period. The operational definitions used to rate each of these symptoms as present or absent during the interview were based on the MDS-derived definitions (HCFA, 1999).
Two independent raters documented the total number and presence or absence of depressive symptoms based on observation during the interview for a subset of 43 participants (total number: r =.76, p <.001; presence or absence of one or more symptoms: =.45, p <.001, respectively). Data for the subsample of participants who completed two GDS interviews also showed good stability for the expression of behavioral symptoms of depression during the interview (n = 34, r =.64, p <.001).
Direct observations were also conducted hourly from 7 a.m. to 7 p.m. on one of the 12-hr assessment days to score one clinical indicator related to psychosocial prevention and intervention (Table 1, Indicator 9), which was defined based on the ACOVE Project (Shekelle et al., 2001; Wenger & Shekelle, 2001). The scoring rule for group-activity participation excluded participants who were bedfast (i.e., MDS item G6a is checked), comatose (i.e., MDS item B1 = 1), or severely cognitively impaired (i.e., MDS item B3e is checked indicative of a recall score = 0).
|
Medical Record Review
A total of nine clinical indicators were modified from a larger medical record review protocol (Saliba, Cadogan, & Roth, 2002). The definitions, scoring rules, and pass rates for each indicator are displayed in Table 1. The pass rate is the proportion of participants who met the pass scoring rule criteria (defined in the second column) divided by the proportion of participants for whom the indicator was applicable (defined in the first column) and could be scored as either pass or fail. The measures based on medical record review (see Table 1, Data Source, medical record) were operationalized according to the ACOVE Project, which identified specific care processes that were determined by expert consensus to be valid (related to a clinical outcome) and feasible to implement by nursing home staff (Saliba et al., 2002; Shekelle et al., 2001; Wenger & Shekelle, 2001). Interrater reliability for scoring (pass/fail) the nine medical record review indicators had values from.55 to 1.0 (p <.05). A maximum of 20 participants had a complete medical record review conducted per facility because of cost issues. Participants who had a physician-recorded chart diagnosis of depression, a current order for an antidepressant medication, and/or an MDS-RAP for disturbed mood were targeted for medical record review to maximize the sample sizes for scoring each indicator related to the appropriateness of treatment. All medical records were reviewed for up to 12 months prior to the review date or the participant's date of admission into the nursing home, if fewer than 12 months.
|
Results |
|---|
|
TOPAbstractDesign and MethodsResultsDiscussionReferences |
|---|
|
2 = 10.63, p <.01). Based on the subsamples with medical record review, there was no difference between the participants on the proportion admitted to the nursing home with a physician-recorded medical record diagnosis of depression. The proportions with a new depression diagnosis after admission or a prescription for a routine antidepressant medication at the time of the study were also comparable between the subsamples with medical record review in the two groups of nursing homes (see Table 2).
Results of Interview
Table 3 shows the prevalence of depressive symptoms according to independent assessments with a standardized interview (GDS-short). Comparable proportions of participants were approached for interview because they had an MDS recall score of 2 (Simmons & Schnelle, 2001), and few participants (n = 17) refused the interview among these subgroups. There were no significant differences in the prevalence rates of depressive symptoms according to independent assessments (resident interview and observation) between nursing homes that scored in the lower versus upper quartiles on the MDS depression quality indicator. Both groups of nursing homes showed a substantial proportion of participants with an MDS recall score of 2 who completed the GDS interview and had a total score indicative of probable depression (GDS of >5: 46% and 41%), though the frequency of behavioral symptoms exhibited during the interview was low (less than one, on average) in both groups. There was also not a significant difference between the lower and upper quartile nursing homes for the proportion of participants who had probable depression according to the GDS and an MDS-RAP for disturbed mood (42% and 48%, respectively).
|
5 (43% vs. 17%, respectively; 2 = 19.12, p <.001). The frequency of behavioral symptoms observed during the interview was also significantly greater among those with a GDS total score above 5 (1.1 ± 1.9 vs. 0.3 ± 0.9, respectively; t = –4.13, p <.001). Participants with probable depression according to the GDS also spent more time in bed (36% ± 28% vs. 29% ± 27%; t = –1.87, p =.062) and less time engaged in any type of activity during the day (34% ± 23% vs. 42% ± 23%; t = 2.65, p <.01) according to hourly direct observation data (7 a.m. to 7 p.m.) compared with those without probable depression according to the GDS.
Care-Process Differences Between Low and High Depression Homes
Homes with low prevalence rates of depression performed significantly better (i.e., higher percentage pass) on two indicators based on medical record review that were physician driven and related to assessment (see Table 1, Indicators 1 and 3). Specifically, a significantly greater proportion of participants in the lower quartile nursing homes had medical record documentation by a licensed provider of a mood assessment (59% vs. 40%; 2 = 7.59, p <.01). Participants in the lower quartile nursing homes who received a new depression diagnosis were significantly more likely to also have medical record documentation of the presence or absence of suicidal ideation (26% vs. 0%; 2 = 6.12, p <.05). No other medical record–based indicators (all shown in Table 1) significantly differentiated nursing homes in the lower versus upper quartile groups. In fact, most nursing homes in both groups had low pass rates for all of the remaining medical record–based indicators, with the exception of the avoidance of specific drugs as the first- or second-line antidepressant treatment (97% and 100% pass rates in lower versus upper quartile nursing homes, respectively). Both groups of nursing homes performed poorly on multiple medical record–based indicators related to the completion of related (i.e., hearing and vision) assessments, documentation of depressive symptoms, and evaluations of a resident's response to treatment (see Table 1).
There were no significant differences between the lower and upper quartile nursing homes on the direct observation measures related to time in bed or engagement during the day (7 a.m. to 7 p.m.). Specifically, percentage of day spent in bed (lower quartile: 30% ± 26%; upper quartile: 33% ± 29%) and engagement in organized group activities (see Table 1, Indicator 9) or other activities during the day (34% ± 22% vs. 31% ± 27%) showed no differences between the two groups of nursing homes.
|
Discussion |
|---|
|
TOPAbstractDesign and MethodsResultsDiscussionReferences |
|---|
The generalizability of these results may be limited by the differential recruitment rate between the upper and lower quartile nursing homes and the low consent rate of the residents across all nursing homes. A higher proportion of nursing homes in the lower quartile group participated in this study compared with the upper quartile group (48% vs. 25%, respectively). However, we found no significant differences on facility characteristics between participating and nonparticipating homes in either group. The overall consent rate for study participants was 35%, and we were unable to compare resident characteristics between participating and nonparticipating individuals. We believe that the low consent rate was due primarily to the narrow timeframe within which residents had to be consented (<2 weeks) and the need to contact and obtain written, informed consent from a designated responsible party for the majority (68%) of residents. Inability to contact a proxy was the primary reason for lack of study participation, as opposed to a formal refusal. Even with these limitations, the study results remain consistent with previous studies, which have shown that depressive symptoms are often undetected and thus untreated among nursing home residents (Abrams et al., 1992; Parmelee et al., 1989; Rovner et al., 1991; Schnelle et al., 2001).
Two critical steps may be necessary to improve the detection and treatment of depression in the long-term-care population. First, nursing home staff should be given specific guidelines for selecting nursing home residents appropriate for interview with a standardized depression screening instrument such as the GDS-short. The validity of the GDS was supported by the results of this study, which showed an association between the GDS and multiple behavioral symptoms indicative of depression (e.g., time spent in bed, engagement in activities during the day). The MDS-based resident-selection criteria (recall score of 2) described in this and previous studies (Simmons & Schnelle, 2001) should guide nursing home staff in their selection of residents appropriate for interview. Second, assessment and treatment care processes recommended in depression-care practice guidelines should be systematically audited and feedback provided to licensed nursing home staff and physicians who are responsible for implementing these care processes. The standardized clinical indicator scoring protocols (see Table 1) described in this study provide the objective technology to accomplish this goal.
|
Footnotes |
|---|
1 Borun Center for Gerontological Research, Department of Geriatrics, School of Medicine, University of California, Los Angeles.
2 Los Angeles Jewish Home for the Aging, Reseda, California.
3 Sepulveda Geriatric Research, Education, and Clinical Center, Veterans Administration Greater Los Angeles Healthcare System, California.
Decision Editor: David E. Biegel, PhD
Received for publication April 3, 2003. Accepted for publication December 1, 2003.
|
References |
|---|
|
TOPAbstractDesign and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  B. M. Bates-Jensen, S. F. Simmons, J. F. Schnelle, and C. Alessi Evaluating the Accuracy of Minimum Data Set Bed-Mobility Ratings Against Independent Performance Assessments: Systematic Error and Directions for Improvement Gerontologist, December 1, 2005; 45(6): 731 - 738. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
