{"id":8861,"date":"2020-10-16T10:37:39","date_gmt":"2020-10-16T10:37:39","guid":{"rendered":"https:\/\/dataethics.eu\/?p=8861"},"modified":"2020-10-16T10:38:41","modified_gmt":"2020-10-16T10:38:41","slug":"we-can-predict-who-will-need-home-care-but-should-we","status":"publish","type":"post","link":"https:\/\/dataethics.eu\/da\/we-can-predict-who-will-need-home-care-but-should-we\/","title":{"rendered":"We Can Predict Who Will Need Home Care – But Should We?"},"content":{"rendered":"

In Denmark, when you turn 65, <\/span>you\u2019re<\/span>\u00a0supposed to get a visit from the municipality<\/span>\u00a0<\/span><\/i>if you need it<\/i>. The purpose is to assess your living situation and find out<\/span>\u00a0if you need<\/span>\u00a0any<\/span>\u00a0<\/span>help, be it\u00a0<\/span>home care\u00a0<\/span>or preventive programs\u00a0<\/span>from the public authorities. When you turn 75 (70 if\u00a0<\/span>you live alone<\/span>) you get the visit no matter what, but at 65 you will have to need it. But how will a municipality know that you need it? It is very expensive to send out a\u00a0<\/span>municipality officer<\/span>\u00a0<\/span>to every home\u00a0<\/span>of a<\/span>\u00a0person, who has just celebrated\u00a0<\/span>their<\/span>\u00a0<\/span>65 year<\/span>s<\/span>\u00a0birthday. And with the perspective of fewer public res<\/span>ource<\/span>s<\/span>\u00a0and a growing elder population, it gets even harder.<\/span>\u00a0<\/span><\/p>\n

\u00a0<\/span>Could data and machine learning help here?<\/span><\/i><\/p>\n

\"\"<\/p>\n

Sasmita\u00a0Kusumastuti,\u00a0<\/span>an assistant professor at the Section of Epidemiology, Department of Public Health and Center for Healthy Aging at the University of Copenhagen<\/span>, believes so.\u00a0<\/span>\u00a0<\/span><\/p>\n

\u201cIn my project we have proven that we can use data and machine learning to develop a predictive algorithm to help the municipality identify, who will be most prone to needing homecare,\u201d\u00a0<\/span><\/i>she says.\u00a0<\/span>\u201c<\/span><\/i>The end goal is\u00a0<\/span><\/i>of course to help more citizens by making p<\/span><\/i>reventive care more sustainable and cheaper in the long run.<\/span><\/i>\u201d<\/span><\/i>\u00a0<\/span><\/p>\n

Anonymized<\/span>\u00a0register data from Statistics Denmark, health data from\u00a0<\/span>Sundhedsdatastyrelsen<\/span>\u00a0and data from Copenhagen Municipality are\u00a0<\/span>the ingredients<\/span>\u00a0in the predictive algorithm.\u00a0<\/span>No data<\/span>\u00a0derives<\/span>\u00a0directly from patients.\u00a0<\/span>It is data such as\u00a0<\/span>living situation<\/span>\u00a0education, family structure and economic status, whether you have been hospitalized,\u00a0<\/span>diagnoses and\u00a0<\/span>medicin<\/span>e<\/span>\u00a0intake<\/span>, and what kind of care has been received. Data goes back<\/span>\u00a0over<\/span>\u00a015 years.\u00a0<\/span>\u00a0<\/span><\/p>\n

\u201cWe study the patterns of events in your life. For\u00a0example\u00a0a male, age\u00a0<\/span><\/i>75<\/span><\/i>\u00a0with high social and economic status, who\u00a0<\/span><\/i>was<\/span><\/i>\u00a0hospitalized some years ago, may be\u00a0<\/span><\/i>due to a<\/span><\/i>\u00a0broken hip, and he has maybe received some medications. All these patterns are used in our predictions which may<\/span><\/i>\u00a0<\/span><\/i>end up saying that there is 80% risk of needing home care\u00a0in\u00a0<\/span><\/i>the near future<\/span><\/i>\u00a0for a person with these characteristics<\/span><\/i>,\u201d<\/span><\/i>\u00a0explains\u00a0Kusumastuti<\/span>.<\/span>\u00a0<\/span><\/p>\n

T<\/span>he project\u00a0<\/span>also look<\/span>s<\/span>\u00a0at historic data\u00a0<\/span>on<\/span>\u00a0those<\/span>\u00a0who already\u00a0<\/span>received<\/span>\u00a0home\u00a0<\/span>care<\/span>\u00a0to\u00a0<\/span>assess\u00a0<\/span>common characteristics which make people prone<\/span>\u00a0<\/span>to<\/span>\u00a0needing homecare<\/span>.\u00a0<\/span>\u00a0<\/span><\/p>\n

This could be implemented in a way that\u00a0<\/span>when you turn 65<\/span>,\u00a0<\/span>and the algorithm shows that<\/span>\u00a0you have a 60% or\u00a0<\/span>higher<\/span>\u00a0risk of needing home care<\/span>\u00a0in the coming years<\/span>,<\/span>\u00a0then you will be offered to be visited by the municipality,<\/span> she explains. She says that the algorithm <\/span>developed in collaboration with machine learning experts\u00a0<\/span>has\u00a0<\/span>a pretty good<\/span>\u00a0accuracy.<\/span>\u00a0<\/span><\/p>\n

\u201cWe made a prediction on\u00a0<\/span><\/i>over one<\/span><\/i>\u00a0million older citizens on historic\u00a0<\/span><\/i>anonymized<\/span><\/i>\u00a0data. We had a learning period between 1997 and 2014<\/span><\/i>, where we analyzed the sequence of events<\/span><\/i>\u00a0or patterns<\/span><\/i>\u00a0<\/span><\/i>in your life e.g. moving to a new place, getting a divorced, living alone, being hospitalized, taking medications etc<\/span><\/i>.\u00a0<\/span><\/i>Using this information, w<\/span><\/i>e predicted<\/span><\/i>\u00a0who would need home care\u00a0<\/span><\/i>at least\u00a0<\/span><\/i>one hour a month<\/span><\/i>\u00a0or admitted to nursing home<\/span><\/i>, and we landed at an accuracy between 85 and 95 percent,\u201d\u00a0<\/span><\/i>she says.\u00a0<\/span>\u00a0<\/span><\/p>\n

The algorithm for now is only\u00a0<\/span>used for<\/span>\u00a0research and has not been\u00a0<\/span>implemented<\/span>\u00a0for real<\/span>\u00a0<\/span>life<\/span>. Today, the municipality make decisions on who needs home care and how much<\/span>\u00a0care<\/span>\u00a0based on questionaries they\u00a0<\/span>have to<\/span>\u00a0fill out for every\u00a0<\/span>preventive home\u00a0<\/span>visit. They\u00a0<\/span>definitely visit<\/span>\u00a0you, when you turn 75 (70 if living alone), if you say yes to the offer.\u00a0<\/span>There has been some reports that som<\/span>e<\/span>\u00a0citizens\u00a0<\/span>or residents\u00a0<\/span>that are ethnically not Danish<\/span>\u00a0<\/span>sa<\/span>id<\/span>\u00a0no to\u00a0<\/span>the home\u00a0<\/span>visit<\/span>\u00a0off<\/span>er<\/span>, but most<\/span>\u00a0say yes. But it is\u00a0<\/span>very hard<\/span>\u00a0to find out who are in need at 65 and thus hard for a municipality to live up to the promises of the welfare law<\/span>\u00a0and optimize their prevention strategies<\/span>.\u00a0<\/span>\u00a0<\/span><\/p>\n

Many Ethical Aspects<\/span><\/b>\u00a0<\/span><\/p>\n

Sasmita Kusumastuti is very aware of the controversial aspects of her project.<\/span>\u00a0<\/span><\/p>\n

\u201c<\/span><\/i>Like a knife or any kind of tool, whether it is good or bad depends on the one operating the tool. Therefore, we need to call attention to data ethics to prevent chaos. We have to ask ourselves: What is right, just, and appropriat<\/span><\/i>e,\u201d<\/span><\/i> she tells Danish IT<\/a>. <\/span><\/p>\n

Among the ethical questions, that\u00a0<\/span>Sasmita<\/span>\u00a0<\/span>is<\/span>\u00a0asking herself, are:<\/span>\u00a0<\/span><\/p>\n