This disclosure relates generally to patient monitoring. Particularly to alarm generation and accompanying artefact and false alarm rejection.
When a patient is being monitored by with multiple parameters, like respiration (Resp), oxygen saturation (SpO2) and ECG (often ST-segment depression/elevation), the combined parameter set usually generates frequent false alarm from the monitored parameters individually, due to e.g. patient motion. Especially Resp is prone to generate false alarms whether it is derived by a bed or mattress motion sensor or through an impedance measurement from the ECG electrodes. A patient may move off the mattress sensor or breathe with his belly (with electrodes on the chest).
Multiple mechanisms to decrease the number of false alarms have been proposed, many trading off sensitivity to specificity. [3] and [4] are examples of the extensive prior art on the apnea and SpO2 monitoring technologies. [6] describes combining information from motion data and SpO2,
The present invention is a novel approach in reducing false alarms without unduly compromising or trading off sensitivity for specificity. It is based on realizing that in many monitoring situations there is a typical evolution of a dangerous sequence of events. Thus an initial patient adverse state leads to one or more, progressively more dangerous states, enabling a logic following this evolution to be specific in generating the alarm, without alarming on individual patient adverse states of the evolution.
One preferred embodiment of this alarm logic is the evolution of sleep apnea into cardiac failure. A patient, often obese and intoxicated, is sleeping. Heavy snoring evolves into apneic episodes as the patient stops breathing for tens of seconds. As the apneic episodes grow longer, the blood oxygen saturation starts to show drops from its normal values (SpO2 drops under 90%). As the dropping SpO2 reaches critical range (typ. <80%, patient dependent). The cardiac muscle starts suffering from oxygen deprivation which frequently is seen as an ST-segment change, or as arrhythmias. Severe arrhythmias relevant here are e.g. Ventricular Tachycardia (VT), Ventricular Fibrillation (VFib) and Asystole. Other arrhythmias like Atrial Fibrillation can also be classified as severe for most patients.
Taken separately the three individual patient adverse states are prone to artefacts due to patient movements. Thus traditional sleep apnea monitors are plagued by false alarms to the extent that the alarming function often is suppressed. Making the alarm generation (e.g. priority escalation) dependent of the event sequence reduces the number of false alarms compared with alarming on limit violation of the individual parameters.
Of course the individual parameters can be made to alarm individually, but for the specific patient group these alarms can be set to be quite insensitive in order to keep the false alarm rate low.
The set of alarming parameters and events can obviously belong to other physiologic state evolutions without limiting the applicability of the present invention.
Note that similar parameters, e.g. Respiration Rate 101 from the bed sensor and Respiration Rate 302 from the chest impedance sensor can be used by the algorithms, either together or alternatively, should the other sensor fail or contain noise. The definition of Apnea should here be understood in the wider sense of severe respiratory insufficiency like very shallow breathing, or very low respiratory rate.
The sequence of events to generate an alarm is defined as a combination of the subalarms of the individual parameters. Subalarm here means a triggered individual alarm that is hidden, and only used as a part of the logic to generate the final alarm. The subalarms usually have their individual limits and priority escalation rules (ref 1). Priorities are low→medium→high, here represented by 1, 2, 3. The final alarm can then be calculated e.g. as
Priority(Final)=(Priority(1)+A*Alarmtime(1)*Priority(2)+B*Alarmtime(2)*Priority(3)) *normalizing factor,
where A and B depend on the parameters and Alarmtime( ) is the time the previous parameter in the sequence has been subalarming. The normalizing factor brings the final alarm priority into the normal range 1 . . . 3. The alarms (not subalarms, but traditional individual alarms) for the individual parameters (e.g. Respiration rate, SpO2 and ST segment) can be set to much more insensitive setting so as again to avoid false alarms, by e.g. using wider limits or longer alarm activation and escalation delays. The term “subalarm logic” is defined here to include alarm limits, activation delay, alarm priority escalation rules and other rules described in [1].
Typically for these sleep apnea traditional individual alarm settings would be as follows: The priority escalation for apnea only would be priority 1 (“note”) reached after 20 s of apnea, with 60 s resulting in priority 2 (“warning”). The maximum priority for apnea could also be “1” to eliminate alarms due to the patient having moved to a position where the respiration sensor would not function.
The individual priority escalation for SpO2 would be more complex, depending both on the time and extent of the limit violation, such that e.g. SpO2<75% would trigger an immediate priority 3 alarm. According to the new logic a high priority could then be reached as SpO2<85% for 2 min following an apnea sustained for >30 s. The ST segment deviation would alarm individually when being <−4 or >+4 mm for 10 min with low priority (“1”), but following an SpO2 low alarm the priority could reach 2 after ST<−2 or >+2 mm for 5 min.
The alarm function for sleep apnea is both to arouse the patient and to alert family or clinicians. A specialized alarm version would be just an indicator in a Holter recording, directing the reviewing clinician's attention to the relevant part of the recording. This saves clinician time and reduces the probability of missing significant events.
The concept of a predefined sequence of subalarms can also be used to set subalarm logic parameters for one step from the values of the previous parameter in the sequence before it subalarmed. Thus the system would detect changes in the patient state from the pre-alarm state. Among these are automatic setting of subalarm limits in sequence; the SpO2 limit may be the pre-apnea SpO2 averaged over e.g. 1 minute minus 5% SpO2. Similarly the ST baseline could be the pre-apnea 5 minute average, and deviations from this value would be subalarmed on.
The details of the escalation of the final alarm is only indicated here; there are obviously several ways of combining the chain of subalarms or events (factors or statuses affecting the alarm logic without being alarms themselves, e.g. “noise”, or “audio pause”—activated) the to produce the final alarm (“sequence logic”).
Examples of sequential adverse events covered by the present invention are:
[1] Alarm generation method for patient monitoring, physiological monitoring apparatus and computer program product for a physiological monitoring apparatus.
[2] Method, Device and Computer Program Product for Monitoring Patients Receiving Care.
[3] Body-worn system for continuously monitoring a patient's bp, hr, spo2, rr, temperature, and motion; also describes specific monitors for apnea, asy, vtac, vfib, and ‘bed sore’ index.
[4] System and method for SPO2 instability detection and quantification
[5] Alarm system that processes both motion and vital signs using specific heuristic rules and thresholds.
[6] Alarm system that processes both motion and vital signs using specific heuristic rules and thresholds.
[7] Lindberg et. al., “Evolution of Sleep Apnea Syndrome in Sleepy Snorers”, American Journal of Respiratory and Critical Care Medicine, Vol. 159, No. 6 (1999), pp. 2024-2027.
Number | Date | Country | |
---|---|---|---|
62109116 | Jan 2015 | US |