Economic evaluation of the PAtient Centered Team (PACT)

Description

This project is a part of a larger research project that evaluates the use of the Patient Centered Team (PACT). The PACT has been established as part of ordinary health services to the elderly patients with complex needs and/or chronic conditions at UNN Tromsø and UNN Harstad. The PACT is interdisciplinary (doctors, nurses, therapists, and pharmacists) and includes employees from UNN and the municipal health service. The intentions with PACT is to develop a better and more extensive health service to the elderly patients and to reduce health care cost.

Goals

The main objective in this sub-study is to analyse the impact of the PACT model on hospital costs, and extra lifetime gained (survival time).

Method

For details on the matched before and after study design see https://ehealthresearch.no/en/projects/pact-i-and-ii

The cost and effect data have been collected, summarised, and compared over a 6-month follow-up period. Baseline costs were collected over a period of 6 months prior to start of the intervention. Two main hospital cost components have been included: the costs related to the PACT model (technology and personnel costs) and the cost related to health services use (hospital inpatient, outpatient, day hospital and emergency services). We used activity-based macro costing to calculate the hospital costs on an individual level recommended by the Norwegian health authorities. The main cost categories included were costs related to the operating room, intensive care, radiology, dialysis, radiation, chemotherapy, and ward costs together with medication and durables (i.e., implants), investment and overhead. The cost of the PACT intervention was included in these cost estimates. For the outcome data, we collected date of inclusion and date of death for each patient and calculated time alive up until 6 months follow-up. From this life-days and life-years gained were calculated.

In the main study, not all intervention patients had an emergency hospital stay at inclusion. Some of the intervention patients were living at home, some had home based care, some were in a municipal institution, and some were included while hospitalised, either planned or as an emergency admission. For this sub-study, we have only included intervention patients with an emergency hospital stay at inclusion since all control patients had an emergency stay at inclusion. This creates more comparable groups for the cost-effectiveness analysis. 1312 patients met the inclusion criteria and were included (aged ≥ 60 years, had a hospital stay at inclusion and had provided informed consent). This left 656 in the intervention and 656 in the control group for analysis.

The participants were consecutively included in the study during the index hospital stay. Intervention patients became eligible when they were referred to the PACT intervention by a health professional, while controls became eligible when matched to an intervention patient during an emergency hospital admission. Therefore, the number of days in hospitals prior to inclusion varied between the included participants. Most of the patients (98%) in the intervention group had days in hospital before inclusion in the study (baseline), while only 26% of the patients in the control group had days in hospital prior to inclusion. The mean (SD) days in hospital prior to inclusion for the intervention and the control group were 7.6 (11.0) and 1,8 (5.7), respectively.

The presence of different number of days in hospital prior to inclusion introduces a selection bias, both for the survival time and the costs. The intervention patients had more days in hospital leading to more patients having survived the acute phase of illness potentially making them healthier than the control group. The control patients have been included in the study closer to the index admission and are therefore more likely to be acutely ill and more at risk of dying in the beginning of the intervention period.

This difference will also affect the costs as acute hospital costs have been shifted from the after-cost to the before-cost estimates. All costs associated with an emergency admission (tests, examinations, interventions, and for some patients even surgery) will be included in the before costs for 98% of the intervention group and only 26% of the control group. This will overestimate the baseline costs and underestimate the post-intervention costs more so for the intervention group.

Data have been analysed using independent sample t-tests, univariate linear model (analysis of covariance) and Kaplan Meier survival curves. Linear model are robust and recommended for analysing cost data. Age at inclusion, cost per patient at baseline and other measures that had a potential to affect costs and survival have been adjusted for in the analyses as covariates (confounders). For the resource use variables, we have only included the total costs at baseline as covariates in the models since days in hospital prior to the intervention, other hospital bed days, readmissions, emergencies, DRG weight and other variables measuring resource utilisation are the basis for and included in the total before cost estimates. A p-value ≤ 0.05 was considered statistically significant.

Results: Assuming we can compare the costs and survival for these two groups, the result shows that PACT-intervention is more effective (10 days gained) and more costly (NOK 35 000 per patient). In an effort to reduce the selection bias, a sensitivity analysis including two models has been conducted. These models showed that PACT did not result in significant changes in neither survival nor hospital costs.

Conclusion

This cost-effectiveness evaluation was not able to establish that the PACT-intervention is cost effective compared to usual care. The results, however, points towards the intervention being more costly and more effective (survival). Correcting for the selection bias by reducing the difference in days in hospital prior to the intervention between the groups would most likely have made the intervention more costly an equally effective. The main reason for this assumption is that if the healthy bias was reduced, this will make the survival effect less for the intervention group. It will further shift the cost of the days in hospital prior to the intervention from before to after and make the intervention even more costly.