CF : Lung TX part 2

[Guest guest]

Health service utilisation

Information about hospital admissions for the study period and the

previous 12 months was collected for all patients from South Australian

Department of Human Services hospital separation data to enable an

intention-to-treat analysis. Patients can be identified across hospitals

and admissions in more than 90% of cases through Medicare numbers.

Unplanned admissions were coded as a separate field in this data source,

and respiratory admissions were identified by Australian National

Diagnosis Related Groups16 176 (pulmonary oedoma and respiratory failure)

or 177 (chronic obstructive airways disease).

Medical Benefits Scheme (MBS), Pharmaceutical Benefits Scheme (PBS) and

Department of Veterans' Affairs (DVA) data were obtained from the Health

Insurance Commission (HIC).17 Each participant received an explanation of

the rationale for collecting these data, and, if agreeable, filled out a

consent form for data release. Information regarding deaths was obtained

from the SA Death Register through the Department of Human Services.

Statistical methods

proportional hazards regressions (adjusted for age and previous

hospitalisation) were used to check for differences in the length of time

in the trial between groups, and to check for differences in time until

death between groups.

For the baseline demographic and prior health service utilisation data

and the baseline quality-of-life characteristics, unadjusted comparisons

between groups were performed using ?2 test, t test or Wilcoxon rank-sum

test, as applicable. Changes in SF-36 component summary scores were

compared between groups using multiple linear regression on a subset of

participants who completed both SF-36 questionnaires, adjusting for five

potential confounders (MBS expenditure and hospitalisation in the 12

months before the study, age, sex, and smoking).

Improvements in functionality were assessed by counting the number of

people in each group considered better, the same or worse than baseline

at the end of the study, with adjustment for potential confounders (the

same five as for the SF-36 analysis) using a multiple ordered logit

model.15

For both the SF-36 and improvements-in-functionality analysis, only

people who completed baseline and follow-up questionnaires were included.

Consequently, adjustment for length of time in the study was not

required.

Differences in hospitalisation for any reason, respiratory admission and

unplanned admission were examined using multiple logistic regression,

adjusting for the same five confounders. Analysis was on an

intention-to-treat basis. Length of time in the study was not a

significant predictor and so was omitted. Average length of stay was

analysed using a t test, including only those patients who had been

hospitalised.

All analyses were completed using STATA statistical software.18

The national evaluators of the coordinated care trials performed power

calculations on a trial-wide basis based on potential changes to SF-36

scores. For our study, power calculations19 were performed for both

quality of life and hospitalisation. For a change of 10 points on the

100-point scale of the SF-36 component summary scores (SD, 10; power,

90%; a = 0.05, using a two-sided t test and an intervention : control

ratio of 2:1) a sample size of 51 (34 intervention; 17 control) would be

required. For a 50% reduction in the incidence of hospitalisation over a

12-month period from a baseline of 42% (power, 90%; a = 0.05, using a ?2

test for an intervention : control ratio of 2:1) the sample size required

would be 245 (163 intervention; 82 control).

Health costs

Costs associated with healthcare service utilisation (MBS and PBS

services, inpatient private and public hospital use, domiciliary care,

and district nursing) and coordinated care are reported as costs per

patient-year. Patient contribution to the cost of services was not

included, and cost data from the DVA and hospital outpatients were not

available. Data were recalibrated to allow for historical difference

between groups during the two financial years 1995 to 1997. Financial

data for the two years were obtained for all study participants and used

to standardise intervention and control subjects at baseline. This

standardisation produced an individual recalibration factor for each type

of service. Also, in adjusting for historical costs, it allowed for

pre-baseline differences such as patient age. Inpatient costs were

estimated using the casemix cost-weighting system, with outliers included

at full cost. The casemix costs were verified in a subsample as

consistent with actual hospital costs.

Results

The Western Respiratory Project involved 223 intervention patients, 154

comparison patients, 92 care coordinator GPs, and six service

coordinators (Box 1). The difference in follow-up time was not

significant (hazard ratio, 0.89; 95% CI, 0.64–1.25). Thirty patients had

less than 90 days' follow-up (intervention, 25; comparison, 5) due to an

adjustment being made to the study start date after recruitment (these

patients were not included in the analysis). The intervention period

varied between patients because of an extended recruitment period with an

associated shortening of the period available for the intervention. One

hundred and eighty-two subjects did not complete the study (Box 1). The

difference in death rates between the groups was not significant (hazard

ratio, 0.62; 95% CI, 0.29–1.28) when adjusted for age and previous

hospital admissions.

Baseline characteristics

At entry to the study, the median age of the intervention group was 10

years older than the geographic comparison group (P < 0.001). The

intervention group had a lower proportion of women (45%, compared with

62% in the comparison group), were less likely to smoke (P = 0.014), less

likely to speak English at home (P < 0.001), and had higher rates of

hospitalisation in the previous 12 months (P < 0.001). In addition, the

intervention group had worse SF-36 physical component summary scores (P <

0.001).

About 90% of participants completed each quality-of-life questionnaire at

the beginning of the study (Box 1). The lowest response rate of 71% was

attained in the intervention group for the SF-36. These non-respondents

were more likely to withdraw from the study, less likely to have had

eight GP visits in the previous 12 months, and more likely to have been

born outside Australia and not to speak English at home. Similar

characteristics were demonstrated by non-respondents for each of the

questionnaires.

The baseline COOP function charts indicated that the intervention group

were more likely to report diminished functioning (physical condition,

breathlessness) and a poorer perception of their overall health and

quality of life. Half the intervention group required assistance in at

least one task of daily living measured by OARS. There was no overall

difference between intervention and comparison groups in activities of

daily living measured by the MBI.

Health services utilisation

For patients receiving coordinated care there was no difference in the

odds of hospitalisation (odds ratio [OR], 1.13; 95% CI, 0.72–1.75),

respiratory hospitalisations (OR, 0.71; 95% CI, 0.40–1.28) or unplanned

hospitalisations (OR, 0.78; 95% CI, 0.47–1.30) after adjusting for

baseline characteristic differences. Length of stay did not differ

between the two groups.

Functionality and quality of life

Changes in quality of life and functionality scores are summarised in Box

2. Multivariate analysis showed no significant difference in change in

SF-36 physical component score. The overall mental component score

improved with coordinated care.

The intervention group experienced less deterioration in two out of three

symptom-related COOP items (emotional condition and pain) and an

improvement in COOP perceived quality of life. There was no difference

between the groups in respect to functional COOP or OARS items.

Healthcare costs

On average, a person receiving coordinated care in the Western

Respiratory Study incurred $8312 per year, including an initial $40

enrolment cost, compared with an average of $6882 per year for a person

receiving usual care (Box 3). Modifying cost outliers to two standard

deviations from the mean did not lead to any significant change in the

results for healthcare costs compared with including outliers at full

cost. This demonstrates that outliers had little impact.

Discussion

We studied the effects of coordinated care in people with chronic and

complex respiratory disease, and found a reduced deterioration in mental

aspects of quality of life, symptoms of pain and emotional condition, but

no difference in physical aspects or functional measures. We found no

cost saving to the healthcare system and no reduction in hospital

admissions.

Key limitations of the study include the lack of a study sampling frame.

Non-participants are not recorded and caution must be taken in

generalising the results. Further, our geographical comparison group was

not as similar to the intervention group as intended, requiring many

adjustments in the analyses for confounders such as age and previous

hospitalisation. The higher prevalence of older men, who were less likely

to speak English at home, suggested that some GPs in the intervention

region had patients of a considerably different demographic background to

subjects recruited by GPs in the comparison region. This study design,

and the extent of associated adjustments, may have weakened the validity

of the differences in outcomes.

The action research approach and lack of interviewer blinding may have

limited scientific rigour. Although this project was a compromise between

scientific evaluation and action research, it provided an important,

large scale, community-based and shared healthcare provider intervention.

Out of necessity, the design was flexible to adapt to the requests of

multiple stakeholders, particularly GPs and their patients, for the

duration of the study. Without such flexibility, implementation across

two broad geographical regions would have been impractical, and it is

doubtful that GPs would have agreed to initiate or maintain involvement

in the study.

Although the study had sufficient power to show differences in admissions

and quality of life, we found no reduction in admissions, and only modest

positive benefit with respect to quality of life. This might be primarily

due to further limitations of the study. Firstly, factors such as the

variation in the period of the intervention and the high dropout rate

reduced the potential for demonstrating an effect. Such factors are not

unexpected in studies of elderly, chronically ill subjects in community

settings. Secondly, owing to patient de-identification in the study

database, we were unable to link patients and GPs in the comparison

group. This meant that we were unable to adjust for clustering by GP in

the analysis. However, as most GPs had only 1–3 patients (intervention

223 patients, 92 GPs; comparison 154 patients, 70 GPs), the clustering

effects are expected to be minor.

The increased dropout rate immediately after 12 months in the

intervention group appears to have resulted from dissatisfaction with the

interview questionnaire. This is an important consideration when lengthy

quality-of-life and other measures are being asked of often frail,

elderly, chronically ill participants.21

Becki

YOUR FAVORITE LilGooberGirl

YOUNGLUNG EMAIL SUPPORT LIST

www.topica.com/lists/younglung

Pediatric Interstitial Lung Disease Society

http://groups.yahoo.com/group/InterstitialLung_Kids/