Preliminary Detection of Acute Exacerbation of Lobar Pneumonia and Heart Failure Using an Anomaly-Detection System Based on a Circadian Rhythm Model Constructed from Non-contact Vital Data
Article information
Abstract
Many frail older patients with multiple comorbidities are hospitalized in long-term care hospitals and nursing facilities. Due to pre-existing conditions and immunosuppressive states, there are significant individual differences, such as weakness, sluggishness, and asymptomatic status. These differences make it challenging to find a patient’s exacerbation using a conventional threshold of vital signs. We developed a circadian rhythm anomaly-detection system designed for each patient, which compares each patient’s past 2 weeks average respiratory rate circadian rhythm with that of the last 24 hours. Respiratory rate was measured using a piezoelectric sensor located under the mattress. Prior to the doctor’s diagnosis of acute exacerbation of lobar pneumonia and heart failure, a bedridden 88-year-old female patient with multiple chronic diseases showed abnormal circadian rhythm status. However, there were no significant changes in vital signs. A circadian rhythm anomaly-detection system appears promising as a future tool to promote medical checkups for the older person.
INTRODUCTION
The number of older people is increasing worldwide. In Japan, many frail older people with multiple comorbidities are hospitalized in chronic care wards and nursing homes, some of whom experience sudden changes in their conditions. In addition, a nationwide shortage of nurses, caregivers, and other staff members makes it difficult to monitor all patients’ conditions, highlighting the need to develop a suitable monitoring system.1) A patient’s condition is monitored by measuring their vital signs; however, these demonstrate significant inter-individual differences, such as weakness, slowness, and asymptomatic state, due to the patient’s pre-existing diseases2) and immunosuppression.3) These differences cannot be accounted for by employing a uniform vital threshold.
Heart failure causes central sleep apnea with Cheyne-Stokes respiration, disrupting the circadian rhythm of the respiratory rate (RR),4) while lung inflammation itself also disrupts the circadian rhythm of the RR.5)
We developed a circadian rhythm anomaly-detection system tailored for each patient. This system compares each patient’s average respiratory rate circadian rhythm over the past 2 weeks with that of the last 24 hours.
Here, we report the use of a circadian rhythm anomaly-detection system based on a circadian rhythm model created from the measured RR that was applied in older people and which predicted the early onset of heart failure and lobar pneumonia in a patient with no changes in conventional vital threshold .
CASE DESCRIPTION
The patient was an 88-year-old bedridden woman living in a nursing home. Her underlying conditions included high blood pressure, chronic atrial fibrillation, chronic heart failure, and a thoracic aortic aneurysm.
The circadian rhythm anomaly-detection system consists of four methods. Long-term vital data were obtained using a piezoelectric sensor (VS1; Konica Minolta Co., Tokyo, Japan) placed under the patient’s mattress to detect slight movements in bed without imposing any burden on the patient Fig. 1A. The RR (1 Hz) was obtained from the micro-movement waveform by optimizing the frequency that captured the characteristics.6) This sensor also acquires body movement information. If body movement is detected, it is excluded from the calculation as a discrete value.6) The obtained RR was then evaluated clinically and was shown to reflect nurses’ conventional visual measurements more closely than other medical devices for monitoring RR.7) A circadian rhythm model was then created automatically for each patient based on the hourly arithmetic average of data from the past 24 hours to 2 weeks. A polynomial approximation was used to estimate the periodicity of the circadian rhythm in Fig. 1B. The periodicity of the data for each day was also compared with the circadian rhythm model as an exclusion criterion for model learning. If there was no periodicity, it was determined that there was no circadian rhythm, and the data were automatically excluded as abnormal. The residuals in the created circadian rhythm model were calculated hourly from vital output data for the last 24 hours, with larger residuals indicating a more-disrupted circadian rhythm Fig. 1C.
(A) Respiratory rate (RR) measurement device and respiratory wave. (B) Circadian rhythm model construction flow. (C) Rhythm status output flow. (D) Discrimination method: a circadian rhythm model was created using only RR.
The rhythm status was created using only the RR and was then used as an index of RR abnormality. Based on a report that an increase in RR of 4 or more bpm within the normal range predicts clinical deterioration,8) a rhythm status of 4 or more was considered an abnormal threshold in Fig. 1D.
The progress of this case and the patient’s vital signs are shown in Table 1 and Fig. 2. The patient’s vital signs, including blood pressure, pulse, temperature, and oxygen saturation, had been measured twice a week by a medical staff member up to February 3, 2023, with no abnormal vital signs or physical conditions other than tachycardia. From February 4th, the rhythm state exceeded the threshold of 4, and on February 5, it increased to 2.4 times the normal level, showing a suspected circadian rhythm disturbance. In response to this change, a nurse examined the patient’s SpO2 level, which was 91%–93% (room air). Her breath sounds in the right lung were weakened, and she was therefore examined immediately by a doctor. On-site portable X-rays and other tests showed cardiomegaly, bilateral pleural effusion, and decreased transparency of the right lung field. The patient was accordingly transferred to an acute care hospital for further examination and treatment and was diagnosed with worsening chronic heart failure and right lobar pneumonia.
Patient’s medical records
(A) Nursing records from 2023/1/15 to 2023/2/6. (B) Respiratory rate by device from 2023/1/15 to 2023/2/6. (C) Rhythm output value from 2023/1/15 to 2023/2/6. (D) The simple arithmetic average from 2023/1/15 to 2023/2/2 confirmed that this bedridden patient had a circadian rhythm; comparing the simple arithmetic average from 2023/2/3 to 2023/2/6 inferred that the respiratory rate remained high throughout the day and night from 2023/2/3 onwards and that the circadian rhythm (CR) was disrupted.
This study was approved by the Ethics Committee of Konica Minolta Inc. (No. 2016-12). Informed consent was obtained from all subjects involved in the study.
DISCUSSION
Currently, many long-term hospitals and nursing facilities rely on human experience to monitor patients. As one reason, this is because monitoring using vital signs in acute care wards (e.g., National Early Warning Score9)) is not efficient because many patients require individualized care. In this study, we used the quantification of daily rhythm changes as an indicator, based on the experience of monitoring in the field. The current circadian rhythm based on RR10) can be used to build a system to robustly detect abnormalities in many patients. This system enabled us to predict the exacerbation of heart failure and lobar pneumonia at an early stage in a patient with no changes in conventional vital sign thresholds. This system can be constructed using any indicator with a circadian rhythm,5,11) such as heart rate or autonomic nerve activity.12-14)
Although we did not visually monitor the patient's respiratory by checking the micro-movement waveforms, we believe that the circadian rhythm of the RR is not problematic. By analyzing the micro-movement waveforms during an increase in RR, shallow tachypnea can be inferred. In addition, waveforms that appear to be Cheyne-Stokes respiration were confirmed two to three times per hour at night.
The circadian rhythm model is always automatically created using data from the past 2 weeks from the current day, but even when the number of calculation days was reduced or compared with past data, the results were close to those of the normal circadian rhythm model. It is believed that normal circadian rhythms always have a constant rhythm (Fig. 2D).
In older person care facilities with no on-site doctors, it may be difficult to evaluate a patient’s condition while taking account of individual differences. By utilizing this system, doctors and medical staff will be able to remotely monitor older people in many facilities. The current system may, therefore, aid staff to assess a patient’s condition, thus preventing hospital-acquired infections and potentially preparing for family members to attend at the time of death. It is essential to accumulate more case data on the use of this system in the future.
Notes
The authors are grateful to the facilities and staff who provided us with the facility data. They also thank Ms. Saeko Nozawa and Shohei Sato for their contribution to the preparation and revision of the manuscript.
CONFLICT OF INTEREST
The researchers claim no conflicts of interest.
FUNDING
None.
AUTHOR CONTRIBUTIONS
Conceptualization, TK, KH; Methodology, TK, TM; Software, TK; Validation, TK, KH; Formal analysis, TK; Investigation, TK; Resources, TK; Data curation, TK, KH; Writing-original draft preparation, TK, KH; Writing-review and editing, TK, KH, TM; Visualization, TK, KH; Supervision, KH; Project administration, TK.