A perfected mismatch: Are prevention and recovery care services really comparable to acute inpatient units?

A recent matched pairs analysis of Prevention and Recovery Care Services (PARCS) concludes that they offer a less restrictive and cost-effective alternative to inpatient facilities for eligible patients (Farhall et al., 2021). This study explored whether PARCS recovery-orientated care model admissions led to better outcomes, i.e., less subsequent 24-hour care requirements and less Inpatient Treatment Orders (ITOs) or Community Treatment Orders (CTOs). However, we observe the two patient cohorts may have initially differed in terms of clinical risk, despite the attempted matching process. Thus, it is debatable whether acute beds can be effectively substituted by subacute residential beds, such as PARCS (Allison et al., 2014). A review by the Victorian Department of Health and Human Services (DoH), published in 2016 found that the ‘use of hospital-based services has not decreased with the introduction of PARCS’ and that PARCS had lower than expected occupancy rates, averaging only 72% between 2009/2010 and 2013/2014 (Department of Health and Human Services, 2016).

Farhall et al. (2021) used propensity score matching (PSM), a method utilised to analyse an observational dataset by setting a priori selected matched variables approximating a randomised experiment. However, PSM has been criticised on the basis that it yields an experimental design with lower standards than necessary, through generating higher levels of imbalance, model dependence, and bias, as well as failing to use all the information available such as latent variables (King and Nielsen, 2019). The latent variables not included in the study include the critical past history that psychiatrists use to guide decisions about admission: previous suicide attempts, previous aggression, severity of previous episodes, utilisation and frequency of past hospitalisation, treatment resistance, non-adherence and levels of family support. The PARCS study only matched for demographic, diagnostic, current legal status and Health of the Nation Outcome Scales (HONOS).

Since sub-acute units serve clinical populations with lower behavioural risk profile than acute inpatient units (Sutherland et al., 2020), the patient group admitted to PARCS and the matched patients requiring admission to an acute inpatient unit may have had substantively different past histories of acuity and risk, despite matching on the HONOS. Such initial group differences could have resulted in the divergent treatment trajectories, e.g., differing rates of ITOs and CTOs. Therefore, the study was potentially biased in favour of PARCS, as past clinical and risk variables were not matched, and could have been higher among the inpatient cohort (Sutherland et al., 2020). The fact that PARCS had average occupancy rates of 72% compared to inpatient unit occupancy rates of 91% (between 2009/2010 and 2013/2014) further suggests that the ‘matched’ inpatient sample in this study likely had more clinical and risk complexity than the PARCS population (Department of Health and Human Services, 2016).

In the 12 months following index admission in 2014, the PARCS cohort had significantly increased number of admissions (0.89 vs 0.62) and duration of admissions (12.24 vs 10.00) compared to inpatient facilities, with no difference in the average number of community mental health care contacts. Therefore, we conclude that PARCS did not prevent 24-hour care requirements. This increase in 24-hour care admissions and duration, explains why PARCS did not demonstrate lower 12-month costs, despite lower bed-day costs. For index admissions, patients treated in acute inpatient units demonstrated greater HONOS score reductions compared to PARCS patients. Accordingly, acute inpatient units were able to deliver better clinical outcomes despite equivalent 12-month costs, thereby delivering potentially better cost-benefit outcomes (measured through HONOS reduction). Increased 24-hour supervised care following an index PARCS admission is not a positive longitudinal recovery-based outcome. The greater HONOS reduction on index admission in acute inpatient units may partially explain the reduced 12-month, 24-hour supervised care need.

In addition, when economically operationalizing the outcomes, the study accepted the DoH standard unit cost as accurate for PARCS bed days ($451 in 2014–2015), but questioned the DoH estimates for inpatient bed days ($688), based on a Victorian Auditor-General’s Office (VAGO) report indicating that the full cost was $1109. For this reason, the sensitivity analysis using the Auditor General’s cost estimates increased the inpatient bed day rate by around 40%. For internal consistency, the study should have used either the DoH cost estimates for both types of admission (PARCS and acute inpatient) or VAGO estimates for the full cost of both types of admission.

In conclusion, we highlight that the PARCS study may have been initially biased in favour of PARCS, by failing to match important past clinical and risk-related variables through the selective PSM process. Increased 12-month, 24-hour care needs following PARCS index admission and lower HONOS index admission reduction suggests that PARCS may not be as preventive or as clinically effective as acute inpatient units. Furthermore, the economic analysis of outcomes used inconsistent cost estimation. Therefore, we conclude that acute inpatient units remain the clinically efficacious standard for acute care, and that PARCS has not longitudinally demonstrated superior clinical or economic outcomes for subacute residential beds.