Article Text

Original research
Proportion of medical admissions that may be hospitalised at home and their service utilisation patterns: a single-centre, descriptive retrospective cohort study in Singapore
  1. Stephanie Q Ko1,
  2. Zhemin Wang2,
  3. Samuel Li Earn Goh1,
  4. John T Y Soong1
  1. 1Division of Advanced Internal Medicine, Department of Medicine, National University Hospital, Singapore
  2. 2Department of Medicine, Alexandra Hospital, Singapore
  1. Correspondence to Dr Stephanie Q Ko; stephanie_ko{at}


Objectives For eligible patient groups, hospital-at-home (HaH) programmes have been shown to deliver equivalent patient outcomes with cost reduction compared with standard care. This study aims to establish a benchmark of inpatient admissions that could potentially be substituted by HaH services.

Design Descriptive retrospective cohort study.

Setting Academic tertiary hospital in Singapore.

Participants 124 253 medical admissions over 20 months (January 2016 to August 2017).

Primary and secondary outcome measures The primary measure was the proportion of hospitalised patients who may be eligible for HaH, based on eligibility criteria adapted for the Singapore context. The secondary measures were the utilisation patterns and outcomes of these patients.

Results Applying generalised eligibility criteria to the retrospective dataset showed that 53.0% of 124 253 medical admissions fitted the eligibility criteria for HaH based on administrative data. 46.8% of such patients had a length of stay <48 hours (‘short-stay’) and 53.1% had a length of stay ≥48 hours (‘medium-stay’). The mortality rate and the 30-day readmission rate were lower in the ‘short-stay’ cohort (0.6%, 12.8%) compared with the ‘medium-stay’ cohort (0.7%, 20.3%). The key services used by both groups were: parenteral drug administration, blood investigations, imaging procedures and consultations with allied health professionals.

Conclusions Up to 53.0% of medical admissions receive care elements that HaH programmes could provide. Applying estimates of functional limitations and patient preferences, we propose a target of ~18% of inpatient medical admissions to be substituted by HaH services. The methodology adopted in this paper is a reproducible approach to characterise potential patients and service utilisation requirements when developing such programmes.

  • Organisation of health services
  • Hospitalization
  • Quality in health care
  • GENERAL MEDICINE (see Internal Medicine)
  • Health Services for the Aged

Data availability statement

No data are available. No additional data are available.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • This study uses existing administrative datasets to analyse service utilisation for patients eligible for hospital-at-home (HaH) retrospectively and could inform future clinical service planning and clinical studies related to HaH.

  • However, contextual factors affecting a patient’s eligibility for HaH, including functional and social considerations, are not readily interrogated in administrative datasets.

  • The study also does not account for patient preference for the mode of treatment, which would affect the acceptance rate.


Hospital-at-home (HaH) is an inpatient bed substitution model that has gained traction recently. Recent focus has been fuelled by shortages of inpatient beds diverted to patients with COVID-19 during surges of the pandemic.1 HaH programmes comprise either or both of two models: an admission avoidance model,2 where patients are admitted into HaH directly from the emergency department (ED), or an early discharge model,3 where patients are admitted from inpatient settings to complete the rest of their treatment at home. Hospital-level care is then delivered to the home of such patients, which includes services such as parenteral medication administration, radiology and other medical procedures. Today, as the surges from COVID-19 stabilise, hospitals continue to examine alternative strategies to inpatient hospitalisation for stable acute medical conditions, of which HaH may be a viable and cost-effective alternative.

Existing HaH programmes have been trialled in different countries with positive results. Multiple studies across Australia,4 5 Europe6 7 and the USA,8 9 and subsequent meta-analyses2 3 10 have shown that for a select group of patients, HaH programmes may be associated with shorter lengths of stay, reduced readmission rates and better patient experiences,3 at reduced costs compared with standard care.3 8 9 Despite emerging evidence, an audit of HaH programmes in Australia shows that HaH only cares for around 3.7% of all inpatient episodes, which accounts for 5% of all bed-days nationally.11 Whether this benchmark is optimal is yet to be established.

Establishing a benchmark of inpatient bed stays that could be substituted by HaH is critical for health services planning new and existing HaH programmes. A benchmark can address adoption, volume and resource allocation to implement HaH successfully. Therefore, this study aimed to apply generalised eligibility criteria to an administrative dataset in Singapore to determine the proportion of eligible patients and explore their characteristics and service utilisation patterns during hospital admission.


Study population

The hospital administrative dataset is obtained from National University Hospital (NUH), one of three tertiary care centres in Singapore. This dataset comprises demographic, medical records, billing and laboratory data.

We included all adults at least 21 years old, admitted into NUH between January 2016 and August 2017 and had full completeness of data. We selected a period unaffected by COVID-19, reflective of general medical admissions and before introducing HaH to the hospital. We excluded patients with a length of stay (LOS) greater than 14 days as they were hypothesised to have non-clinical factors for continued inpatient care (eg, social issues, or need for long-term inpatient rehabilitation). The Strengthening the Reporting of Observational Studies in Epidemiology checklist was used to ensure this descriptive cohort study meets international reporting requirements for observational studies.

Defining HaH eligibility criteria

We reviewed published studies on HaH programmes to extract common eligibility criteria5 6 8 9 12–16 (Appendix A in online supplemental file). Nine HaH studies published between 2001 and 2020 were found to have well-described eligibility criteria (Appendix A in online supplemental file). Seven of these studies8 9 12–16 excluded patients who were deemed clinically unstable, five studies9 13–16 explicitly listed the conditions constituting clinical instability and three studies6 14 15 excluded patients with terminal illnesses with life-expectancy of 6–12 months. The studies also excluded patients who were anticipated to require diagnostic or therapeutic procedures exclusive to hospital settings,8 15 patients who had severe psychiatric disorders,13 14 active pulmonary tuberculosis14 or were undergoing haemodialysis.8 14–16 Non-clinical exclusion criteria included residence outside the hospital catchment area,9 10 12 13 absence of social support,14–16 sub-optimal home environments,8 9 residence in institution care5 8 13 14 and the inability to give informed consent.6

The study team selected final eligibility criteria based on three principles: (1) applicability to a general medical HaH service in Singapore, (2) no restriction based on diagnosis group and (3) adaptability to variables available in the administrative dataset. The final eligibility criteria selected by the study team are summarised in table 1.

Table 1

Final HaH exclusion criteria

Determining the proportion eligible for HaH

The generalised exclusion criteria were applied to the study population sequentially to determine the volume of patients eligible for HaH care. Eligible patients were then categorised according to their LOS, ‘short-stay’ for LOS less than 48 hours and ‘medium-stay’ for LOS between 48 hours and 14 days. The cut-off was chosen to distinguish patients with very short LOS who may potentially benefit from an admission-avoidance HaH model (ie, admission to HaH direct from the ED), or with a longer LOS that may potentially benefit from early-discharge HaH models (ie, admission to HaH from inpatient wards).

Baseline demographics and clinical outcomes collected

Baseline demographics collected included age, gender, ethnicity, Charlson Comorbidity Index (CCI),17 diagnosis and whether they were admitted to the acute medical unit, a short-stay ward. Diagnosis codes were collapsed from the International Statistical Classification of Diseases and Related Health Problems 10th Revision to Healthcare Cost and Utilisation Project Clinical Classifications Software.18 Clinical outcomes of interest were the LOS, mortality and 30-day readmission post-discharge.

Service utilisation patterns of HaH-eligible patients

Common elements of inpatient care were identified and quantified to characterise service utilisation patterns in each HaH group. This includes administration of parenteral drug therapy, laboratory or imaging procedures and consultations by allied health professionals. Prescription data were used to determine the proportion of patients requiring parenteral medication administration. Such medications included fluids, antibiotics, insulin (subcutaneous), diuretics and electrolytes. Billing data were used to determine the proportion of patients who had received common imaging investigations such as X-rays, CT scans, MRI and ultrasound scans. Similarly, the proportion of patients reviewed by allied health professionals (physiotherapists, occupational therapists, speech therapists and dieticians) was also quantified using billing data.

Statistical analysis

In this descriptive cohort study, categorical variables were described with proportions and percentages. For continuous variables, those that were normally distributed were described with means and SD and those that were not were described with medians and interquartile ranges. A subgroup analysis was done for patients in the ‘medium-stay cohort’ to further subset into patients with LOS 48 hours to 7 days and 8–14 days. All data analysis was done using R V.1.1.383.

Patient and public involvement



Proportion of patients eligible for HaH

Of 124 253 acute hospital admissions to NUH from January 2016 to August 2017, 80 692 (64.9%) were admissions to the Department of Medicine. After applying the final eligibility criteria (figure 1), 42 732 (53.0%) of all medical admissions received elements of care that HaH could provide. Of these, 20 011 patients (46.8%) were classified as ‘short-stay’ admissions (LOS up to 48 hours) potentially benefiting from admission avoidance HaH, and 22 721 (53.1%) were classified as ‘medium-stay’ admissions (LOS between 48 hours and 14 days) potentially benefiting from early discharge HaH.

Figure 1

Study cohort selection. ICU, intensive care unit; LOS, length of stay.

Characteristics of patients eligible for HaH

Of the 42 732 patients who may be eligible for HaH, patients in the ‘short-stay’ cohort tended to be younger (median age 64, IQR 51–76) than the‘medium-stay’ cohort (median age 69, IQR 57–81) (table 2). ‘Short-stay’ patients had a slightly lower comorbidity burden (median CCI 4, IQR 2–5) than the ‘medium-stay’ patients (median CCI 4, IQR 3–7). The ethnic and gender distribution was similar in both groups. The majority of patients were admitted to the general medicine specialty.

Table 2

Demographics and clinical characteristics of HaH-eligible patients

The top diagnoses differed between both groups. Non-specific chest pain, intestinal infections, diabetes with complications, pneumonia and upper respiratory tract infections were the five most frequent diagnoses in the ‘short-stay’ cohort. Urinary tract infections, pneumonia, diabetes with complications, intestinal infections and skin and soft tissue infections were the five most frequent diagnoses in the ‘medium-stay’ cohort.

Clinical outcomes of patients eligible for HaH

In-hospital mortality was 0.6% in the ‘short-stay’ cohort and 0.7% in the ‘medium-stay’ cohort. Unexpected mortality rates, defined as cases referred to the coroner, were low in both groups (<0.1%) (table 2).

Median LOS was 1.28 days (IQR 0.83–1.71) in the ‘short-stay’ cohort and 3.83 days (IQR 2.82–5.74) in the ‘medium-stay’ cohort. 30-day readmission rates were lower in the ‘short-stay’ cohort (12.8%) compared with the ‘medium-stay’ cohort (20.3%).

A subgroup analysis of the ‘medium-stay’ cohort into patients who stayed less than 7 days (n=19 496) and those who stayed between 8 and 14 days (n=3225) revealed that patients who stayed more than 7 days had the highest readmission rate (23.8%) and mortality rate (2.1%) (Appendix B in online supplemental file).

Service utilisation patterns of HaH-eligible patients

For administration of parenteral medications in both cohorts, intravenous fluids were given in 28.4% and 47.5% of the ‘short-stay’ cohort and ‘medium-stay’ cohort, respectively, and intravenous antibiotics in 12.1% and 35.6% (table 3) .The most common antibiotics administered were ceftriaxone, co-amoxiclav and piperacillin-tazobactam(Appendix C in online supplemental file). 76.1% of all doses of parenteral medications were administered after office hours (defined as 17:00 to 08:00) (Appendix C in online supplemental file).

Table 3

Service utilisation patterns of HaH-eligible patients

Patients in the ‘medium-stay’ cohort used more procedures than the ‘short-stay’ cohort. More than 95% of patients in both groups required blood tests. X-rays were used in 22.0% of the ‘short-stay’ cohort and 40.5% of the ‘medium-stay’ cohort. 7.6% and 5.6% of the ‘short-stay’ cohort required MRI and CT scans, respectively, compared with 13.9% and 13.9% of the ‘medium-stay’ cohort.

Consultations with allied health professionals were used less frequently in the ‘short-stay’ cohort compared with those in the ‘medium-stay’ cohort. 21.5% of patients in the ‘short-stay’ cohort required physiotherapy, compared with 51.6% in the ‘medium-stay’ cohort. Other commonly consulted allied health professionals included occupational therapists, dietitians and to a lesser extent, speech therapists and podiatrists. Estimates of workforce requirements for allied health professionals are shown in Appendix C in online supplemental file.


Our retrospective cohort study of inpatient administrative data shows that 53% of all medical admissions in a large tertiary hospital received elements of care that could be provided by a comprehensive HaH programme, that is, would be clinically eligible for HaH. However, for patients to eventually be hospitalised at home, apart from clinical eligibility two other criteria must be fulfilled apart from clinical eligibility—both functional suitability and patient consent. Applying an estimate from previous studies of functional suitability of 62%19 and patient agreement rate of 56%,20 we estimate that ~18% may be a suitable benchmark for HaH programmes to target for substitution of medical admissions (Appendix D in supplemental file).

To our knowledge, this is the first study that uses administrative data to estimate eligibility and service utilisation patterns for a potential HaH service. The results from this study provide a basis for estimating the demand for a full-scale HaH care model in an urban setting; for Singapore in particular. Extrapolating data from systematic reviews,2 3 10 18% of medical admissions being substituted by HaH would provide substantial advantages to health systems in three ways: (1) improved patient satisfaction, (2) reduced direct costs of care, (3) expand inpatient hospital capacity without having to build additional physical infrastructure; with equivalent or improved LOS, readmissions and mortality.

Clinical service utilisation may also guide estimates for service development. Most patients eligible for HaH use services such as administration of parenteral medicines, blood tests and consultations with allied health professionals. To effectively substitute these services, HaH programme must develop several key attributes. First, out-of-hours nursing must be available, whether through round-the-clock staffing or mobilising community healthcare professionals. Considering that most intravenous therapy was administered after hours, regimens promoting daily or twice daily dosing may need to be uniquely created for the HaH service to function efficiently. Second, advanced imaging facilities must be available to patients, and appropriate transport for patients back to the hospital is provided. Third, social care provisions must be considered to assist with activities of daily living, considering the high utilisation rate of physiotherapy and occupational therapy.

Our findings build on existing evidence regarding eligibility volumes for HaH programmes. A point prevalence study of acute and subacute inpatients in Royal Melbourne Hospital suggested an eligibility rate of 11.1%.19 These differences may be explained by the difficulty estimating functional impairment in administrative datasets and differences in diagnosis groups. Cellulitis, digestive malignancy and respiratory infections were their top 3 most common admissions,11 compared with non-specific chest pain, pneumonia and urinary tract infections as the top 3 in our study. The common admission diagnoses are heterogenous and may be non-specific (eg, chest pain and diabetes with complications). Therefore, patient selection criteria in HaH programmes should be based on excluding patients with treatment needs that cannot be provided at home rather than including selected diagnoses only.

Our study has several limitations. First, the retrospective nature of the large data analysis and variables available in an administrative dataset presents inherent limitations which may affect the benchmark estimate. It excludes clinical judgement at the point of recruitment and assumes that clinical deterioration requiring intensive care unit admission can be predicted prospectively. Furthermore, it cannot account for the variable acceptance rate among different populations, which various social and cultural contexts and healthcare financing models may influence.20 Finally, it cannot effectively identify social and functional limitations that exclude patients from HaH. To address the above, we have applied estimates from existing studies to our proposed benchmark of 18%. Although there may be overlap in the exclusion criteria—for example, patients with functional limitations that render them ineligible for HaH treatment may also wish to be treated in a hospital setting instead, our estimate of 18% would be conservative.

Second, the HaH exclusion criteria we applied in this study may only be generalisable to some HaH programmes. Some HaH programmes can support blood transfusions, provide supplemental oxygen and manage nursing home patients; in such circumstances, the proportion of eligible patients would only increase. Third, the association of <48 hours LOS with admission avoidance and >48 hours LOS with early supported discharge is an arbitrary estimation. In reality, there will be significant overlap between both groups. Fourth, we recognise that some patients deemed eligible for HaH may also be able to receive care provided by alternative or future care models, such as ambulatory emergency care or community care services.21 Fifth, due to limitations in the dataset, readmissions to hospitals apart from our institution would not be captured and may lead to under-reporting of clinical outcomes. Sixth, we did not estimate potential costs for HaH services as financial data were not readily available.


Combining clinical eligibility estimates of 53% of patients from our administrative database with 62% functional suitability and 56% patient agreement from previous studies, we propose a target of 18% of inpatient medical admissions that mainstream HaH services can substitute. Further studies are warranted to validate this benchmark. There may be additional value in developing and validating algorithms to further risk-stratify patients and predict which patients may most benefit from HaH programmes.

Data availability statement

No data are available. No additional data are available.

Ethics statements

Patient consent for publication

Ethics approval

Ethics approval was obtained through the Domain Specific Review Board, Singapore (approval number 2020/00127).


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors SQK, ZW and JTYS developed the concept for the study. SQK, SLEG and ZW conducted the literature review, writing and editing of the manuscript. SQK performed the data analysis for the study. SQK, SLEG, ZW and JTYS were involved in interpreting the results, editing and proofing of the manuscript. SQK is responsible for the overall content as the guarantor.

  • Funding This study was funded by the National Medical Research Council Health Services Research Grant (HSRGMS20nov-0004).

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.