More proactive and targeted care

More proactive and targeted care

Reducing costs through more proactive and targeted care, which allows providers to intervene earlier to keep people well, supported by powerful analytics.

Related content

More proactive targeted care

There is enormous potential to reduce cost by early intervention through more proactive and targeted care. Powerful analytics can be used to spot early warning signs in patients in community and hospital settings and avoid significant number of deaths and admissions. Patient data can be used to predict clinical risk, enabling providers to target resources where they are needed most and target problems that would benefit from early intervention. This is an area where the deployment of technology can rapidly deliver significant improvements in outcomes and savings. It should be a high priority for investment — particularly in a hospital setting. And as genomic information begins to become routinely captured as part of clinical examinations, the importance of analysis of this kind will only become greater.

Pretty soon instead of the patient seeking the hospital, it will be the hospital seeking the patient.

— Jagruti Bhatia, KPMG in India

Predictive analytics

Analysis of electronic datasets has the potential to more accurately predict healthcare demands in the future. Computer-based algorithms, drawing on patients’ clinical and demographic data, can generate risk scores identifying those at higher risk of avoidable readmissions. Northern Arizona Healthcare in the US, for example, has found that sending risk scores to discharge planning nurses reduced emergency readmissions by 45 percent. Use of analytics can extend beyond simple readmission prevention, however, and may have a role in predicting those in the community who are likely to use healthcare services in the near future. So-called ‘case finding’ tools are well established but have been held back in the past by having to rely on limited, out of date and poor quality data. Systems of the future will draw not just on electronic clinical data (which is much cheaper to harvest) but data from home monitoring equipment and even — in some markets — personal data held by retailers or telecoms companies.

Monitoring of vital signs and early identification of those at risk

Remote monitoring technology offers significant potential for reducing avoidable use of health resources and targeting staff time cost-effectively at those most in need. A number of proprietary systems have demonstrated promising results in homes and hospitals — for example, VitalPAC in the UK (see below). In the US, Cerner have developed a system to identify the early symptoms of sepsis — a leading cause of avoidable harm that is often missed. By continuously monitoring key clinical indicators for potentially septic patterns, Cerner estimates its system could reduce in-hospital patient mortality by 24 percent and length of stay by 21 percent, saving US$5,882 per treated patient.

Specific lessons

Implement vital signs monitoring solutions at scale: A number of hospitals have tried to implement vital signs solutions in one or two wards rather than across the hospital as a whole. They found this led to duplication of work and an increased administrative burden in trying to marry separate paper and electronic systems. This suggests that to maximize the benefits of vital signs monitoring, it should be implemented across the whole hospital.

Think carefully about the data that should be used for predictive analytics: Analyses of predictive models for case finding have found that drawing on a higher number of detailed data sets improves accuracy.Systems that use real-time clinical and population-based data sources are likely to be more medically useful for time sensitive interventions than those drawing from retrospective data sets.

Make the most of unstructured data: The vast majority of healthcare data is unstructured (such as doctors and nurses notes) and it will be essential to find ways to make best use of it. This may mean finding technological solutions to convert it into structured data, such as natural language processing and text mining. 

Since there is a large category of people who deteriorate over 2 or 3 days, with a pattern you can pick up… behaviors that are exacerbating the problem or symptoms. If you collect those in a systematic way… Then you can intervene.

— Adam Darkins, Medtronic

Predictive algorithms for readmission (Clalit, Israel)

Clalit is Israel’s largest not-for-profit insurer and provider serving 3.8 million people. It has developed an algorithm for predicting patient readmission which is used for patients admitted to any of its 27 hospitals. In practice this means that clinicians have access to a list of all their patients that have been discharged from any hospital in the country on a daily basis, ranked according to their calculated risk of readmission. They are then able to undertake a process that is already hard-wired into the EHR — phoning the patient, asking them about risk factors and whether they have the drugs and support they need. A study found a 4 percent drop in a 30-day readmission for high-risk patients as a result.2

Vital signs monitoring in practice: VitalPAC, UK

VitalPAC is a technology solution for hospitals with a range of products allowing for electronic monitoring of patients, including VitalPAC Nurse which identifies patients at risk through early warning scores, VitalPAC Doctor which gives mobile access to real-time patient information to improve handovers and task prioritization, and and infection control tool, VitalPAC IPC. Trialled in UK hospitals, VitalPAC has reported significant improvements in outcomes including: 15 percent reduction in mortality; 70 percent reduction in cardiac arrests; 50 percent reduction in unplanned transfers to ITU, 90 percent reduction in norovirus outbreaks and a reduced overall length of stay. The company claims the system has a return on investment of between four and six times.

Footnotes

1Record Data Med Care 2015;53: 283–289)16 Billings J, Georghiou T, Blunt I, et al. (2013) Choosing a model to predict hospital admission: an observational study of new variants of predictive models for case finding. BMJ Open; 3:e003352

2Shadmi, E, Flaks-Manov, N, Hoshen,M., Goldman, O., Bitterman, H., Balicer, R. (2015) Predicting 30-Day Readmissions With Preadmission Electronic Health

Connect with us

 

Request for proposal

 

Submit

KPMG's new digital platform

KPMG's new digital platform