|
Which patients with heart failure respond best to multidisciplinary disease management? |
Riegel B, Carlson B, Glaser D, Hoagland P |
|
|
Record Status This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn. Health technology The intervention evaluated was multidisciplinary team management to promote self-care abilities in patients with heart failure. The intervention included educational materials, counselling, discharge assessment, outpatient support groups, physician collaboration, home visits by specialist nurses, and telephone case management. A multidisciplinary team of pharmacists, dieticians, social workers, heart failure specialty nurses and registered nurses provided the intervention with expertise in heart failure. The intervention was compared to usual care which was defined as management of patients by the primary care physician or cardiologist, patient education by hospital nurses using available educational materials, dietician and social worker contact for problematic situations, and referral to home care for some patients.
Economic study type Cost-effectiveness analysis.
Study population The patient population was patients with chronic heart failure. Patients with transient heart failure, cognitive impairment, primary renal failure requiring dialysis, severe psychiatric illness, inability to speak English, or who were discharged to long term care facilities were excluded.
Setting The intervention and usual care evaluated were provided in a secondary care setting, in five hospitals in Southern California, USA.
Dates to which data relate The intervention was implemented in 1996. The evaluation was completed in 27 months. The authors did not report the start or end dates of the evaluation, so it is not possible to determine the exact dates to which the effectiveness, resource use and cost data relate.
Source of effectiveness data The effectiveness data were derived from a single study.
Link between effectiveness and cost data Resource use and cost data was collected retrospectively for the same sample of patients as the effectiveness data.
Study sample The sample size required was estimated before the study. The power calculation used analysis of variance (ANOVA) for a small to moderate effect size (not specified). This gave a Cohen d statistic =0.35. Combined with 80% power and a 2 tailed significance level of 5%, this gave an estimated sample size of 120 per group for the primary analysis of all patients. The authors used a significance level of 0.10 for the primary sub group analysis. Patients in the intervention group were enrolled when they had an admission to one of the two hospitals with the multidisciplinary disease management programme for heart failure and a confirmed diagnosis of chronic heart failure. The patients enrolled into the intervention group were matched with patients from three hospitals providing usual care. The authors did not report the number of patients who refused to participate in the study prior to enrolment or the number of patients who were excluded from the study.
Study design The evaluation used a non-randomised clinical trial design, with concurrent controls. The patients were matched for pre-admission functional status, co-morbidity and age. Nearest available metric matching techniques were used to find the closest match for each intervention group participant, from the pool of unmatched usual care patients. Patients were followed up for 6 months following discharge from the index hospital admission. The authors reported that 323 patients were enrolled and 240 completed the study. Of the 83 participants who did not complete the study, 22 died (13 intervention group and 9 usual care group) and 11 were admitted to a skilled nursing facility (7 intervention group and 4 usual care group). Of the participants who withdrew voluntarily from the study, 66% were in the usual care group. The withdrawals were older than the final study sample. The differential loss to follow up resulted in a statistically significantly shorter duration of follow up in the intervention group than in the control group (169 days versus 180 days, p=0.03). The authors stated that this was not clinically significant. Treatment allocation was not concealed from study participants, care providers or investigators.
Analysis of effectiveness The analysis of the data collected was based on those participants who completed the scheduled follow up. As noted above, only 240 of the 323 patients enrolled were included in the analysis. One of these patients was an outlier for the measure of total cost and was excluded from the analysis on statistical and clinical grounds. The rationale for this was not provided. The authors did not report the group in which the patient was enrolled. It was not clear whether the authors were evaluating effectiveness in terms of measures that were proxies for health outcome or in terms of measures of resource use for acute care. The primary measure used to assess effectiveness was the percentage of all patients admitted for heart failure during the 3 months following discharge from the index admission. This was chosen as the most sensitive indicator of treatment effectiveness. The authors did not report the rationale for this choice. Other measures of effectiveness reported by the authors were the overall hospitalisation rates at 3 and 6 months, overall days in hospital at 3 and 6 months, heart failure admissions at 6 months, and heart failure related days in hospital at 3 and 6 months. Analysis of variance was used to assess the intervention effects outlined above for the primary analysis for all patients and 4 sub groups. The sub groups were defined by pre-admission functional status measured by the Specific Activity Scale. The authors reported this to be a standardised interview with higher validity and reproducibility than the other functional measure used in the study (New York Heart Association classification method).
There were no statistically significant differences in the demographic characteristics of gender, marital status, education or annual income at enrolment. The usual care group was younger and had a shorter length of stay in the index admission than the intervention group. These differences were statistically significant. There were no statistically significant differences in all but one of the clinical characteristics of the usual care and intervention groups. The serum sodium level of the usual care group was 1 point lower than that of the intervention group at discharge from the index admission. The demographic characteristics of age and length of index admission stay were included in the secondary regression analysis to determine predictors of the 3 month heart failure admission rate and identify sub groups of patients who might benefit most from the intervention.
Effectiveness results The effectiveness results were as follows:
At 3 months there were no statistically significant differences in the overall hospitalisation rate (0.37 intervention versus 0.36 usual care), heart failure hospitalisation rate (0.22 intervention versus 0.52 usual care), overall days in hospital (1.68 intervention versus 1.46 usual care) or heart failure related days in hospital (0.89 intervention versus 0.48 usual care).
At 6 months there were no statistically significant differences in the overall hospitalisation rate (0.63 intervention versus 0.60 usual care), heart failure hospitalisation rate (0.32 intervention versus 0.23 usual care), overall days in hospital (2.66 intervention versus 3.03 usual care) or heart failure related days in hospital (1.31 intervention versus 1.08 usual care).
There were statistically significant differences between the intervention and usual care groups for 2 of the sub groups, at 6 months follow up. These differences were:
for the group with highest pre-admission functional status (normal, Class I) all hospitalisations (0.61 intervention versus 0.21 usual care, p<0.10);
for the group with highest pre-admission functional status (normal, Class I) all days in hospital (3.74 intervention versus 0.79 usual care, p<0.10);
for the group with minimal functional compromise (Class II) all days in hospital (1.72 intervention versus 5.64 usual care, p<0.10).
The secondary regression analysis indicated that the following variables were significant predictors of heart failure admission rates:
higher levels of co morbidity (p=0.002);
better in patient functional capacity (p=0.02);
length of index hospitalisation (p=0.01); and
the interaction between length of index hospital stay and allocation to the intervention group (p=0.02).
The final regression model only explained 17.2% of the variance in heart failure admission at 3 months.
Clinical conclusions The authors concluded that multidisciplinary disease management had no effect in an unselected population of patients with heart failure. The intervention had differential effects on sub groups defined by pre-admission functional status. Those with normal functional status had increased hospital admissions and days in hospital in the intervention group compared with usual care. Those with minimal functional compromise had reduced hospital days in the intervention group. There was no intervention effect for the two groups with moderate to severe functional compromise.
Methods used to derive estimates of effectiveness Methods used to derive estimates of effectiveness were consensus, experts' opinion, and authors' assumptions.
Measure of benefits used in the economic analysis As noted above, it was not clear whether the authors defined effectiveness in terms of measures of, or proxy measures for, clinical outcome or health benefits. The multidisciplinary disease management was shown to have no effect when effectiveness was measured in terms of acute care use. If you, the reader, are satisfied that use of acute health care is a reliable proxy for clinical or health outcomes, then the study may be regarded as a cost minimisation study and no measure of health benefits is necessary.
Direct costs The direct costs included in the analysis were the variable costs of days in hospital and were measured from the records of each patient enrolled in the study. These excluded overhead and running costs. The authors did not report the resource items actually included in the cost per day of hospitalisation. They stated that the variable hospital costs were measured for each patient using a hospital cost accounting database. Direct costs of non-hospital services were not included. The costs were not discounted, which was appropriate for the 6-month time frame of the study. The authors did not report when the resource use data were collected or the price year used.
Statistical analysis of costs The total costs of acute care were analysed for statistically significant differences between the intervention and usual care groups. The analysis was conducted on untransformed data, despite skewed distributions. The authors used analysis of variance and the non-parametric equivalent (Mann Whitney test) and stated that they did not observe differences in the results of the 2 tests. The authors did not state whether the resource use or cost data reported were average values, and if so, whether medians or means were presented. A measure of variance was presented, but the authors did not report which measure was used (e.g. standard error, standard deviation). The total costs were included in the regression analysis described for the effectiveness measures above.
Indirect Costs Indirect costs of lost productivity were not included.
Currency US dollars ($). No currency conversions were reported.
Sensitivity analysis Sensitivity analysis was not conducted.
Estimated benefits used in the economic analysis See effectiveness results above.
Cost results The authors reported the costs of all hospital-based care used in the 6 months following discharge from the index admission for heart failure. The costs of the index admission were not included. It was not clear whether the costs of the intervention which were not provided in the hospital were included. There were no statistically significant differences in total costs at 3 months ($1,369 intervention versus $1,355 usual care) or 6 months ($2,361 intervention versus $2,566 usual care), heart failure related costs at 3 months ($632 intervention versus $317 usual care) or 6 months ($1,024 intervention versus $686 usual care). There were statistically significant differences in: total costs in Class I (normal function) patients ($3,613 intervention versus $930 usual care); total costs in Class II (minimal functional compromise) patients ($1,638 intervention versus $5,114 usual care); and heart failure related costs in Class I patients ($1,696 intervention versus $115 usual care).
Synthesis of costs and benefits The authors did not synthesise the costs and benefits.
Authors' conclusions The authors concluded that multidisciplinary disease management for unselected patients with chronic heart failure did not have an effect on the use or costs of acute health care. They also concluded that the intervention increased the use of health care and costs in patients who were fully functional before hospital admission and that the most benefit of the intervention was for patients with minimal pre-admission functional compromise.
CRD COMMENTARY - Selection of comparators The authors justified the evaluation of disease management in a population of unselected heart failure patients with reference to the lack of evidence in this area. They described alternative methods of disease management to that evaluated (case management and clinic models). The authors did not justify the use of usual care as a comparator for multidisciplinary disease management. You, as a user of this database, should consider whether usual care is a relevant or widely used alternative in your own setting.
Validity of estimate of measure of effectiveness The analysis of effectiveness was based on a non-randomised trial with concurrent, matched controls. As the authors noted the use of a non-randomised design may have compromised the internal validity of the evaluation. The patients in the control group (usual care) were recruited from different hospitals to those in the intervention group. This may have helped to minimise the effects of treatment contamination (i.e. the possibility that the usual care group received similar services to those in the intervention group), which would reduce any differences in effect between the groups. However, it could introduce confounding effects due to unobserved differences between the hospitals or patients. The authors did not report whether the study sample was representative of the study population. The groups were not comparable at enrolment in age or length of index hospital admission. This may have influenced the subsequent use of health care. However, these factors were included in the regression analysis of treatment effect. In addition, only those patients who completed follow up were included in the analysis. Patients who died or were transferred to a skilled nursing facility were excluded. There were more patients withdrawn for this reason in the intervention group than in the usual care group. This may have overestimated the minimal benefits found for the intervention group. Inclusion of these patients may also have changed the conclusion that there were no differences in effect between the intervention and control groups.
Validity of estimate of measure of benefit The authors did not measure clinical outcomes or health benefits. They did not define whether the measures of effect used (use of acute health care) were intended as proxy measures of health outcome. As noted above there were differences in the number of deaths between the groups, which, if included in the analysis, may have affected the conclusions. The measure of effect used did not evaluate the impact of the interventions on patients' health status.
Validity of estimate of costs The authors did not provide a full list of the resource use or price data used to estimate the direct costs included in the analysis. The direct costs were restricted to those used for hospital based care. The direct medical and non medical costs of care provided outside the hospital were not included. This may have underestimated the costs of care. In particular, if the disease management intervention replaced use of non-hospital care with hospital care, the relative costs of usual care may have been underestimated, which would affect the conclusions of the analysis.
Other issues The authors noted that the results of the study were not consistent with previous studies, which had found a benefit for disease management in heart failure patients. The authors suggested that this might have been due to differences in intervention intensity, study design or the patient population used. The authors suggested that the latter was the most likely reason, since previous studies used selected patients most likely to benefit from the intervention. The authors also noted that the study results were consistent with evaluations of disease management in patient populations without heart failure.
As noted above, the authors do not provide sufficient information to judge whether the results are generalisable to alternative patient populations or settings.
Implications of the study The authors suggested that heart failure disease management programmes should only be offered to groups of selected patients who are likely to benefit. They also recommended that additional research is required to validate the findings of their study and to identify other sub groups of patients who may benefit.
Source of funding Supported by Merck & Co Inc, Whitehouse Station, NJ, USA; Sharp HealthCare Foundation, San Diego, California, USA; and the San Diego Chapter of the American Association of Critical Care Nurses, San Diego California, USA.
Bibliographic details Riegel B, Carlson B, Glaser D, Hoagland P. Which patients with heart failure respond best to multidisciplinary disease management? Journal of Cardiac Failure 2000; 6(4): 290-299 Indexing Status Subject indexing assigned by NLM MeSH Activities of Daily Living; Aged; Analysis of Variance; Cost-Benefit Analysis; Disease Management; Female; Heart Failure /classification /therapy; Humans; Male; Patient Care Team /organization & Patient Education as Topic /organization & Patient Selection; Predictive Value of Tests; Prognosis; Program Evaluation; Self Care /economics /methods; Severity of Illness Index; Treatment Outcome; administration; administration AccessionNumber 22001000243 Date bibliographic record published 30/11/2001 Date abstract record published 30/11/2001 |
|
|
|