# Interpretation of Health-Smart Home Data and Implications for Clinical Decision-Making: Inductive Content Analysis

## TL;DR

This study explores how nurses interpret health data from smart homes and finds that better data visualization and training are needed to improve clinical decision-making.

## Contribution

The study introduces a novel inductive content analysis of nurse interpretations of smart home sensor data, highlighting visualization preferences and clinical challenges.

## Key findings

- Nurses identified key patterns in sleep, mobility, and home engagement from sensor data.
- Bar and line graphs were preferred over pie charts for interpreting health-smart home data.
- Unclear sleep metrics and lack of clinical context hindered decision-making.

## Abstract

Health-smart home technologies offer real-time sensor-based monitoring of older adult activities of daily living, allowing for early detection of changes in health. The way clinicians interpret and use this data, particularly in visualized formats, such as bar, line, and pie graphs, remains underexplored.

A qualitative descriptive study design with a quantitative component was used to explore how nurses interpret sensor-derived health data from health-smart homes in 3 cases.

Using an inductive content analysis approach, we analyzed nurses’ qualitative interpretations of existing sensor-derived health data from health-smart homes from 3 older adults living with ambient whole-home sensing. Nurses provided structured written feedback on visualized trends in sensor-derived health data, including activity, sleep, and mobility patterns.

The findings highlight both opportunities and challenges of using sensor-derived health data in older adults’ care. Nurses identified key patterns in sleep, mobility, and home engagement, but interpretation difficulties, such as unclear sleep metrics and lack of clinical context, hindered decision-making. Nurses preferred bar and line graphs over pie charts for interpreting these data. Survey results show a statistically significant difference in how nurses rated different graph types (χ²2=17.1, P<.001), with pie charts rated significantly lower than both bar and line graphs (P<.001 and P=.008, respectively). These findings underscore the need for improved data visualization and integration to enhance the clinical utility of sensor-derived health data from health-smart homes.

Findings indicate that nurses were able to provide accurate interpretations of the sensor-derived health data from health-smart homes. However, there is a need for improved visualization techniques and clinician training to optimize health-smart home data for early intervention. Standardized approaches to data representation could enhance nurses’ ability to detect and act on subtle yet important information about older adults’ health changes occurring in home settings.