From: Selecting long-term care facilities with high use of acute hospitalisations: issues and options
Research question: to … | • Find the fewest facilities to accumulate numbers of hospital events? |
• Identify resident- or facility-level characteristics associated with higher (or lower) event rates so as to inform intervention design? | |
• Find facilities that have high hospital presentation rates even if explained by resident characteristics? | |
• Find facilities that, independently of their facility or resident characteristics, have high event rates? | |
• Find facilities that after adjusting for non-modifiable characteristics, have unexplained high rates of presentations? | |
Hospital event type as endpoint of interest | • All hospital visits, or acute/ED presentations, or acute admissions? |
• All or selected diagnoses only, e.g. those classified as potentially avoidable (PAH)? | |
• If only selected diagnoses, e.g. PAH, were codes predefined or selected/amended after data was gathered? | |
LTC facility type | • Limit to particular facility types – e.g. lower-level care? |
• Use only facilities with complete or near-complete data? | |
• Is distance or time to hospital likely to impact referral decisions? | |
• Use only facilities of a certain size (for power & cost considerations) | |
• Need to stratify by e.g. geography, or match in pairs for randomisation? | |
Resident care type | • Use only long-stay residents, or include short-stayers? |
• Limit to those in certain levels of care, e.g. low-level care, or dementia care, or in one age group, or those with public funding, or those with a particular clinical history? | |
Cohort assembly | • Include all residents at any one time, i.e. cross-sectional? |
Or all entering (or leaving) the facility during a pre-defined period? | |
Or all using the facility at any time during a period? | |
Time period of events | • Hospital events over what time period? |
• Data collected retrospectively or prospectively? | |
• In a special study, or with routine data collection? | |
Adjustments during analysis | • Can results consider person-time, e.g. on death or moving away? |
• Can results consider facility-level characteristics? If so, how? | |
• Can results consider resident-level characteristics? If so, how? | |
Measure for reporting and ranking | • Report a count, a proportion, a rate over time, a facility-related effect size from model, a residual from a fitted statistical model, or a change in rank between two methods? |
• Express as rate per bed, per resident, per resident year, or relative to other facilities, to an earlier report or to a “best practice” target? | |
Data quality, completeness & recency | • What is the extent of missingness in data – facilities, outcomes or data items? |
• Is missing data correlated with particular variables so as to lead to bias? | |
• Are data current, or could changes have occurred since collection? | |
• How reliable are measures, ratings, and coding? |