Semi-automated Patient Monitoring HST950/6.872What is patient monitoring?Who needs intensive monitoring?Reasons for monitoring & storing dataVariety of monitoring situationsWhy admit a patient to the ICU?IT in the hospitalMost important parameters to monitorOther important informationData volume in the ICUMultiple databases, multiple sourcesTelemetry InstrumentationWireless Telemetry InfrastructureAlarms - ICUSubjective scales… and (semi) objective parametersSAPS II: New Simplified Acute Physiology ScoreCritical Care Scoring SystemsBut … can we trust the data? Errors of commissionErrors of omissionFalse alarms & noiseABP:Signal quality assessmentSignal quality of ICU dataFalse ECG arrhythmia alarm frequency in ICUECG Alarm Example 1ECG Alarm Example 2Data fusion for false alarm suppressionAre we over-monitoring?Conclusions on monitoringReadingsThe MIMIC II Database – a resource for all!Clinical studies that might interest you?Outpatient/home monitoringIssues with outpatient monitoring Data heterogeneityLength of stay of ICU patientsReadmission rates in NE USA in 1990Relative Alarm FrequencyDistribution of major problems (ICD9)Statistics of ICU patientsU.S. Health care statistics 2008US Hospital Statistics 2008U.S. DemographicsReasons for admissionHealth Care spending Per Person 2007Heuristics and weighted sums …ICU ECG arrhythmia alarm definitionsSemi-automated Patient Monitoring HST950/6.872Gari Clifford, PhDhttp://www.mit.edu/~gariWhat is patient monitoring?“Repeated or continuous observations or measurements of patients, their physiological function, and the function of life support equipment for the purpose of guiding and assessing the efficacy of patient management decisions, such as therapeutic interventions.”(Gardner and Shabot)• Automated (noise prone)• Semi-automated (decision support, human filters)• Manual (integrative, low-cost, traditional, subjective)(Usually well trained, intelligent monitors!)Who needs intensive monitoring?There are at least 5 categories of patients who require intensive physiological monitoring:1. Patients with unstable physiological regulatory systems (e.g. a patient with impaired respiratory system)2. Patients with suspected life-threatening conditions (e.g. acute MI)3. Patients at risk of developing a life threatening condition (e.g. major post-operative patients)4. Patients in critical physiological state (e.g. trauma/shock)5. Mother and baby during the labor and delivery processReasons for monitoring & storing data• For decision support and management of patient• To allow the review of historical data • To organize and allow concise reporting (billing?)• To allow data mining for retrospective studies• To measure severity of illness and provide optimization of resource allocation• Automated data fusion, and ‘conflict resolution’• To allow quality control through assessment and evaluation of performance (outcomes, LOS, infection rates, costs, etc)Variety of monitoring situations– Paramedics (light/variable)– E.R. (light/variable)– O.R. (hyper-intensive)– I.C.U. (intensive)– General Wards (light)– Outpatient / Home (light)Why admit a patient to the ICU?• Life-threatening acute problems• Requires ‘constant’ vigilance and care (by humans and/or technology)• Patient is unstable• Leave when stable:– Life sustaining tech. not req’d– Continuous round-the-clock care not req’dIT in the hospital• Sensors • Monitors• Data transmission• FeedbackMost important parameters to monitor• RR (respiratory rate), • SpO2(oxygen saturation), • ECG (electrocardiogram), • HR (heart rate), • Core temperature (inside the heart) • Peripheral temperature (on top of the instep), • CI (cardiac output index, CO/m2),• Systematic pressures (SSAP, systematic systolic arterial pressure; SDAP, systematic diastolic arterial pressure; SMAP, systematic mean arterial pressure), • CVP (central venous pressure), • Pulmonary pressures (PSAP, pulmonary systolic arterial pressure; PDAP, pulmonary diastolic arterial pressure; PMAP, pulmonary mean arterial pressure), • svO2(oxygen saturation in the lung artery), • ETCO2 (outcoming carbon dioxide), • FIO (ingoing oxygen), • Diuretics,• Drip drugs (type and rate)• Bolus (oral/hypodermic) drugs • Patient’s weight, • Fluid balance (ingoing and outcoming fluids) • Electroencephalogram (EEG)• Intra-cranial pressure (ICP)Other important information• X-rays (masses, pneumonia, pneumothorax, fracture)• MRI & CT scans (stroke)• Ultrasound diagnostics (fluid, labor)• Admit note/patient history• Nursing progress notes• Problem lists• Alarms• Events/procedures/interventions (intubation, surgery, transfers …)• Laboratory tests• Movement & consciousness (GCS, actigraphy)Data volume in the ICU• 1MB-2GB/patient/dayParameter Amplitude RangeFreq Range (Hz)Sampling Freq (Hz)Bit rate (kb/s)Amplitude resolution(bit)Cuff BP 10-400 mmHg0-60 ~0.001 <10-48IBP 10-400 mmHg0-60 ~125 0.25 8-16PAP 0-50 mmHg0-60 ~125 0.25 8-16CVP 0-50 mmHg0-60 ~125 0.25 8-16ECG ±4mV 0.01-500 100-2k Hz 0.2-6 8-16SpO2 80-100% 0-30 ~80 <0.2 8-10C.O. 4-25 L/min 0-20 ~<0.001 10-58Temp 32-40 oC 0-0.1 1 < 10-38Multiple databases, multiple sources• Practices differ in different units• Data formats differ in each DB• Synchronization difficult• Data may be contradictoryTelemetry InstrumentationWireless Telemetry Infrastructure• Complex• Delays• Time stamp issues• Protocols (HL7)• Security issues• Patient IDAlarms - ICU• Straight thresholds– HR < 60 BPM == Bradycardia– No inspiratory gas flow ≥ 20s == Apnea• Multi-parameter thresholds– No inspiratory gas flow < 20s, bradycardia, cyanosis, or pallor == Apnea– Data Fusion (HR from ECG, BP and PPG)• Subjective/objective reasoning & logic– Pallor (subjective?)– Temporal reasoning – if this, then that, check this, then …• Complex classifiers– Linear regression (SAPS)– Nonlinear/nonstationary(ANN, EKF)Subjective scales• ~80% inter-observer agreementGlasgow Coma ScoreRiker Scale2+2+2≤9 … Patient needs CT scan to rule our traumatic brain injury… and (semi) objective parametersSAPS: Simplified Acute Physiology Score• ƒ(Age, HR, SBP, Temp, Resp, Urine O/P, BUN, HCT, WBC, Glucose, K, Na, HCO3, GCS)HCT = HematocritWBC = White Blood Cell CountGCS = Glasgow Coma Score.BUN = blood urea nitrogen.Urine O/P = total urine output in first 24 hoursSAPS II: New