Promoting resilience to weather-related and seasonal provocations to health in people with multimorbid heart disease: a prospective pragmatic, randomised trial

Randomisation and masking

An independent data management team implemented a blinded, computer-generated randomised protocol. A pre-determined randomisation sequence with block groups by biological sex randomised participants one-to-one to the intervention or standard management groups.

Procedures

Information about socio-demographic status (based on multifaceted, individual profiling), medical history, hospital treatment, and planned post-discharge management were collected for each participant during their index hospitalisation, as previously described.¹³ Profiling specifically focused on their vulnerability to external provocations to health (infectious diseases, direct provocations to cardiovascular health related to weather events) from a bio-behavioural perspective,⁶ including assessment of their behaviour patterns and any history of recurrent hospital admissions during a particular season (defined as more than half of admissions during one season during the preceding twelve months)¹⁰ (Supporting Information, part 1; appendix II). After discharge from hospital, all participants had access to high quality specialist and primary care as part of standard management, including standard hospital avoidance programs, outpatient management by specialist physicians, routine follow-up by primary care physicians, subsidised pharmacological treatment, and referral to allied health care services and multidisciplinary chronic disease management programs, as required. Treating physicians received a copy of the hospital discharge summary, as well as information about the trial (including group allocation).

Intervention group participants received standard management as well as the RESILIENCE program, which entailed more frequent clinical surveillance and support during the twelve months after discharge, according to their assessed vulnerability to external provocations to health (Supporting Information, part 1, figure 4).¹³ Seven to fourteen days after discharge, a qualified nurse with postgraduate training visited the participant at home. A standardised protocol¹³ was used to assess their home environment, behaviours, and clinical status to identify areas of vulnerability to changes in the weather and other external provocations to health from a multifactorial perspective.⁶ The RESILIENCE nurse, a qualified physician, and the participant and their family (when appropriate) reviewed the priorities and individual circumstances of the participant at a dedicated clinic (virtual or in-person, as preferred) within 30 days of hospital discharge. Recommendations for optimising clinical management and promoting resilience were enacted immediately or conveyed in a comprehensive report to the participant's health care team. The RESILIENCE nurse coordinated the additional care and provided individualised support during the 12-month active management period, including several home visits and arranging additional RESILIENCE physician reviews if required.

At twelve months, all living study participants were invited to attend a health review, and a summary of the findings and recommendations were sent to the participant's health care team.

Outcomes

Data for outcome measures during study follow-up (until 28 July 2023, the end of the 12-month follow-up period for the most recently recruited participants) were collected from electronic medical records by investigators blinded to group allocation. The primary outcome was the proportion of days alive and out of hospital during complete follow-up with respect to the maximum number possible (29–972 days). Secondary outcomes were event-free survival (ie, did not reach the composite endpoint of any-cause hospital re-admission or death), timing of hospital re-admission and length of hospital stay with respect to weather events and season, and 12-month change from baseline in health-related quality-of-life (EQ-5D¹⁷). Further pre-planned secondary outcomes, including more detailed analyses of timing of events relative to weather conditions, specific sub-group comparisons (eg. those with and without chronic respiratory disease), and a health economic analysis (if appropriate), will be reported elsewhere.

Statistical analyses

Based on data from a preliminary study of weather-associated health events in cardiac patients,¹⁰ we calculated that 150 participants in each group would provide 85% power (α [two-sided] = 0.05) to detect a 10% change in the primary outcome in the pilot study (0.860 days alive and out of hospital during twelve months’ follow-up; standard deviation [SD], 0.035 days). As a result of the prolonged COVID-19 lockdowns in Melbourne, the recruitment target was revised on 5 May 2022 to 100 participants per group, with extended follow-up to 28 July 2023 (minimum 12-month follow-up).¹³

Our analyses followed a pre-specified statistical analysis plan (Supporting Information, part 1). Discrete variables are summarised as frequencies and proportions, continuous variables as means with standard deviations (SDs) or medians with interquartile ranges (IQRs).

All efficacy analyses were undertaken on an intention-to-treat basis and blinded to group allocation. The statistical significance of differences between the intervention and standard management groups in the proportion of actual days alive and out of hospital was assessed in negative binomial regression analyses, adjusted for length of follow-up. Survival and event-free survival were depicted in Kaplan–Meier survival curves, and differences assessed in Mantel–Cox log-rank analyses.

The statistical significance of between-group differences in the change in EQ-5D scores from baseline to twelve months was evaluated in an independent samples t test. As the difference was not statistically significant, we report only mean change with 95% confidence intervals (CIs).

The timing of hospital re-admissions and length of hospital stay was assessed in multinomial logistic regression analyses. Model fit was assessed using the likelihood ratio and deviance goodness-of-fit tests; the association between baseline characteristics and hospital re-admissions was assessed using multiple logistic regression (backward stepwise method) using the variables listed in Box 2; 95% CIs were derived from model-based standard errors.

Event-free survival and events during specific seasons were assessed in Cox proportional hazards models (entry model: proportional hazards confirmed by visual inspection) and multiple logistic regression (backward stepwise removal [P > 0.1 in univariate analyses]; age, sex, and group randomisation were included as fixed variables, as were initially all variables in table 1 in the Supporting Information and, for events during specific seasons, time of exposure to the season of interest.

All analyses were performed in IBM SPSS Statistics 28.0.0.0 (190).

Ethics approval

The Austin Health human research ethics committee approved the study (HREC/56509/Austin-2019), and all participants provided written informed consent for participation.

RESILIENCE intervention group

During follow-up, 94 of 103 intervention group participants (91%) were visited at home by the RESILIENCE nurse, a median of fourteen days (IQR, 0–23 days) after discharge; 49 people (52%) reported no immediate concerns. Twenty-seven people exhibited poor awareness of weather and climate, seventeen did not have the socio-economic resources to manage areas of concern, and the behaviour of sixteen participants placed them at risk of weather-related cardiovascular events; five people lived in physical conditions that exposed them to weather extremes.

Ninety-one participants attended the RESILIENCE clinic a median of ten days (IQR, 6–15 days) later. All attenders received education and goal-setting advice and agreed to receive weather alerts by text message. More specific measures included adjusting pharmacological therapy (78 participants, 86%), help with applying for home heating or cooling subsidies (42, 46%), organising pertussis (42, 46%), pneumococcal (38, 42%), or varicella (32, 35%) vaccinations, and the commencement of vitamin D supplements (seven, 8%). Vulnerability to external provocations to health was subjectively assessed as being low for 29 participants (32%), moderate for 39 (43%), and high for 23 (25%). Plans regarding areas of concern were communicated to each participant's health care team, and follow-up was tailored to their individual needs and assessed level of risk (sample patient history: Supporting Information, part 2).

Re-admission to hospital and death

A total of 470 hospital admissions and 3874 days of hospital stay during follow-up were recorded for 138 of the 203 trial participants (68%); 38 people (19%) died during follow-up. The days alive and out of hospital proportion (the primary endpoint) was 89.1% (SD, 22.0 percentage points) for the intervention group and 86.5% (SD, 24.7 percentage points) for the standard management group (adjusted difference, 3.04 percentage points; 95% CI, –0.21 to 6.29 percentage points).

Neither the differences in the overall numbers of deaths in the intervention and standard management groups (22 of 103, 21% v 16 of 100, 16%; P = 0.25) nor the composite endpoint of time to first hospital re-admission or death (72, 70% v 72, 72%; P = 0.97) were statistically significant (Box 3). The number of hospital re-admissions was similar for the intervention and standard management groups (224 v 246; adjusted difference, –2.75%; 95% CI, –0.39% to 5.89%), as were the numbers of days in hospital (1834 v 2040; adjusted difference, 2.63%; 95% CI, –1.13% to 6.39%).

Box 3. Survival and event-free survival* for participants in the RESILIENCE trial: Kaplan–Meier survival analyses

CI = confidence interval.

* No hospital re-admissions or death.

In an analysis adjusted for the variables listed in Supporting Information, table 1, several socio-demographic (living alone) and clinical factors (coronary artery disease, heart failure, diabetes, depressive symptoms, reduced kidney function, low vitamin D level, pneumococcal vaccination), but not group randomisation, were associated with increased risk of hospital re-admission or death during follow-up (Supporting Information, table 1).

Quality of life

The EQ-5D was completed at both baseline and at 12-month reviews by 65 standard management and 81 intervention group participants. For the standard management group, mean self-rated health was 67.2 (SD, 18.9) at baseline and 72.9 (SD, 14.6) at twelve months; for the intervention group, the mean rating was 63.1 (SD, 20.6) at baseline and 69.9 (SD, 17.4) at twelve months. The mean between-group difference in the change between assessments was not statistically significant (standard management v intervention: 1.10; 95% CI, –5.58 to 7.79).

Timing of events

The number of hospital re-admissions peaked during April and May (the transitional months from warmer and dryer autumnal weather to markedly colder and wetter winter conditions in Melbourne), both the raw figures and after adjustment for time at risk based on recruitment date, standardised by applying time-exposure weighting to months of follow-up (Box 4).

Box 4. All-cause hospital re-admissions of participants in the RESILIENCE trial, by month of re-admission (A);* mean high and low temperatures in Melbourne, 19 November 2020 – 28 July 2023, and major weather events during the 72 hours preceding re-admissions (B)

* Red bars: summer; teal: autumn; blue: winter; green: spring.

Of 470 hospital re-admissions, 192 (41%) followed weather events during the preceding 72 hours. One hundred of these events were dynamic temperature changes: 48 warming events (peak temperatures rose by a mean 21.0°C [SD, 3.3°C]) and 52 cooling events (peak temperatures dropped by a mean 20.2°C [SD, 2.9°C]); 63 of these temperature events were during summer. There were also 68 rain periods (64 during autumn/winter) and 24 thunderstorms (21 during summer/autumn).

After adjustment for the time of follow-up, the number of days of hospital stay was lower for the intervention than the standard management group in the three summer months of December (136 fewer days; –3.38 [95% CI, –6.06 to –0.71] days/person), January (138 fewer days; –4.26 [95% CI, –6.84 to –1.68] days/person), and February (130 fewer days; –6.10 [95% CI, –8.91 to –3.29] days/person); the overall difference for the three summer months was –4.10 (95% CI, –6.87 to –1.33) days/person. The numbers of days of hospital stay were higher for the intervention than the standard management group for five of twelve months, but the differences were not statistically significant, nor were the between-group differences significant for any other season (Supporting Information, figure 1 and table 2).

Limitations

Apart from the adverse impact of the COVID-19 pandemic on trial recruitment and delivery of the study intervention, the major limitation of our pragmatic trial was that study group allocation was not masked for participants or health teams. We undertook therapeutic measures when clinical need was identified (eg, iron deficiency) during baseline profiling. Standard management was consequently enhanced for both groups, but we did not monitor levels of routine care or adverse effects of the study intervention. The smaller than planned study size reduced our ability to detect differences in hospital re-admission rates. Causes of death were not known, and we did not perform an interaction analysis to determine whether the effect of the intervention differed by season. As a single centre study conducted in the comprehensive Australian health care system and with the specific weather conditions in Melbourne during the COVID-19 pandemic,²⁹ any interpretation or extrapolation of our findings to other health care systems or locations must be cautious. A detailed and lengthy analysis of who might have benefited from the intervention, together with health economic analyses of the cost of different components of health care and supplementary socio-economic support (eg, subsidised heating or cooling) will be reported elsewhere; this information will inform our plans for testing a revised intervention in a more selected group of patients.

Conclusion

A climate-focused, nurse-coordinated, multifaceted intervention was ineffective in increasing the proportion of days alive and out of hospital for people followed up for at least twelve months after hospital admissions with multimorbid heart disease, compared with standard, post-discharge management. However, given the growing number of people with multimorbid heart disease and bio-behavioural vulnerability to climatic challenges to health, strategies that cost-effectively improve health outcomes should be pursued, especially in the face of climate change.³⁰ Careful analyses of the timing of significant weather events indicated that the efficacy of our intervention could be improved by modifying its components and applying it in a more targeted manner to reduce weather-triggered events.