Improv­ing atti­tudes to food: apply­ing behav­iour­al sci­ence to nudge peo­ple into health­i­er dietary choices

Dim­itris Koutoukidis:
Here we are, so thank you very much for invit­ing me. I am Dim­itris, I’m a research fel­low at the Uni­ver­si­ty of Oxford and my research focus­es on food pur­chas­ing as well as devel­op­ing inter­ven­tions for dietary weight loss. And my stud­ies are fund­ed through pub­lic funds from the Nation­al Insti­tute for Health Research. And today I’m going to be talk­ing to you about how we can improve atti­tudes to foods through apply­ing behav­ioral sci­ences and nudges, par­tic­u­lar­ly focus­ing on food pur­chas­ing and pret­ty much fol­low­ing on from Pas­cal’s excel­lent presentation.

Dim­itris Koutoukidis:
So a lit­tle bit of con­text. We know that diet, togeth­er with tobac­co use are the two most impor­tant deter­mi­nants of health. And as in the U.S., the same applies in the UK, where the diet at the pop­u­la­tion lev­el is sub­op­ti­mal as we eat quite a lot of sug­ar, sat­u­rat­ed fats, salt as well as col­ors. Now, try­ing to improve that at the pop­u­la­tion lev­el has been very, very challenging.

Dim­itris Koutoukidis:
And there­fore, we need to know the inter­ven­tions to try and see how we can actu­al­ly man­age that. In this con­text, food pur­chas­ing is a very promis­ing option because it is a key deter­mi­nant of food con­sump­tion. And there­fore that gives us a clear oppor­tu­ni­ty for inter­ven­tion with quite rich.

Dim­itris Koutoukidis:
Now we can sim­plis­ti­cal­ly think of these inter­ven­tions as those that tar­get the indi­vid­ual, and those that tar­get the envi­ron­ment. And both speak­ers have touched upon those. So the indi­vid­ual-lev­el inter­ven­tions such as offer­ing peo­ple a health­i­er alter­na­tive, or a swap required reflec­tion and ana­lyt­i­cal deci­sion-mak­ing. But mean­ing­ful deci­sions about food are not reflec­tive, but instead they’re auto­mat­ic reac­tions prompt­ed by envi­ron­men­tal cues. And there­fore there’s quite a lot of inter­est on envi­ron­men­tal-lev­el inter­ven­tions, such as order­ing the default at the point of choice, the so-called choice archi­tec­ture, or nudg­ing inter­ven­tions. And as [pho­net­ic Gareth 00:02:11] men­tioned, it’s quite chal­leng­ing to run all these inter­ven­tions with the retailers.

Dim­itris Koutoukidis:
So what we did was that we decid­ed to take a step back and try to see if we can see an effect in the lab, in an exper­i­men­tal online set­ting. So the aim of the first study, and I will present two of those today to you. It was to test effec­tive­ness of an indi­vid­ual-lev­el inter­ven­tion and envi­ron­men­tal-lev­el inter­ven­tion on the sat­u­rat­ed fat con­tent of the shop­ping bas­ket dur­ing an online shop­ping experiment.

Dim­itris Koutoukidis:
So, par­tic­i­pants were recruit­ed and they were asked to buy 10 items from a shop­ping list. These were all items that were high in sat­u­rat­ed fat, and for which alter­na­tives with low­er sat­u­rat­ed fats were also exist­ing. And we tried to look for food cat­e­gories that were typ­i­cal in the UK diet. We asked par­tic­i­pants to buy one from each cat­e­go­ry and noth­ing more. And try­ing to buy things that they and their fam­i­ly would typ­i­cal­ly eat.

Dim­itris Koutoukidis:
So what we did was that we recruit­ed 1,240 peo­ple for pro­lif­ic, and we ran­dom­ized them to either be offered a swap, either alter­ing the default order of foods, both inter­ven­tions com­bined, or a con­trol group. In the first inter­ven­tion where we offered peo­ple a swap, peo­ple essen­tial­ly went to our online super­mar­ket where they can see the instruc­tions, they could with­draw at any point. And they could essen­tial­ly look for an item as with any oth­er super­mar­ket, either through the drop­down menu, or by typ­ing on the bar. So if, for exam­ple, some­body want­ed the ched­dar cheese, they will click on that, they will see all the prod­ucts in ran­dom order, and they will click on the one they want­ed to buy. And at that stage, they will get a swap which will be matched for brand, weight and price. And they could choose to either keep these or stick with the orig­i­nal option.

Dim­itris Koutoukidis:
So then this will be added to the trol­ley, and then they will con­tin­ue doing that for the fol­low­ing 10 prod­ucts. And then they’re going to check­out. In the sec­ond inter­ven­tion, where we essen­tial­ly changed the order of prod­uct and we rank them based on the amount of sat­u­rat­ed fat that they had. That meant that essen­tial­ly what [inaudi­ble 00:04:48] was that the prod­ucts were in an ascend­ing order of sat­u­rat­ed fat. So this time when peo­ple clicked for the ched­dar cheese, what you see here is typ­i­cal­ly half of cheeses, lighter cheeses. So in this case, what do you see in start? Par­tic­i­pants, so half of cheeses, and lighter cheeses in pret­ty much in the whole first page. And it was only until the sec­ond page that they were actu­al­ly able to see a real ched­dar cheese. They could, as always, click on these and see the nutri­tion infor­ma­tion as with any oth­er super­mar­ket. They could add things to the trol­ley, go back on the menus and con­tin­ue on with the next product.

Dim­itris Koutoukidis:
Now, we were a lit­tle bit wor­ried that because that was sim­ply an exper­i­ment and peo­ple weren’t actu­al­ly spend­ing their own mon­ey that what they’re going to do is that they’re going to rush through this very quick­ly, click on every sin­gle but­ton that they could think of and try to fin­ish as quick­ly as pos­si­ble. So we don’t have process eval­u­a­tion and maybe we should see that peo­ple spend about 20 min­utes on study, browse about 25 pages and bought about 10 prod­ucts in their bas­kets. So, so far so good.

Dim­itris Koutoukidis:
Our pri­ma­ry out­come was that change in the per­cent­age ener­gy from sat­u­rat­ed fat in the total bas­ket and com­pared to the con­trol group that he had no inter­ven­tion. And he had about 26% of the ener­gy from sat­u­rat­ed fat from their final shop­ping bas­ket. Offer­ing a swap, reduce that by two per­cent­age points, which was sta­tis­ti­cal­ly sig­nif­i­cant and what we had pow­ered our study to detect, because we thought that this was a clin­i­cal­ly rel­e­vant reduc­tion. Alter­ing the order, reduced that by five per­cent­age points and that was sig­nif­i­cant­ly low­er, both com­pared to the con­trol and com­pared to the swaps. And com­bin­ing the inter­ven­tions togeth­er led to a small decrease fur­ther, but this was not sig­nif­i­cant­ly low­er than alter­ing the order. It was only sig­nif­i­cant­ly low­er than swaps and the con­trol group. What this tells us essen­tial­ly, is that both inter­ven­tions were effec­tive alter­ing the order was more effec­tive than swaps. And com­bin­ing the two, essen­tial­ly there was no added ben­e­fit of adding the swaps.

Dim­itris Koutoukidis:
We’re also pleased to see that there was no dif­fer­ence in their cost, which is a major chal­lenge for many of these inter­ven­tions. And there was no sig­nif­i­cant dif­fer­ence by group in the total shop­ping bas­ket. And there was also no dif­fer­ence in terms of sex, age, eth­nic group, BMI, edu­ca­tion, or income.

Dim­itris Koutoukidis:
Now mov­ing on to the sec­ond study, of course, offer­ing the swaps was less effec­tive than alter­ing the order. But many super­mar­kets do see this as a much more viable first step approach. So we want­ed to explore a lit­tle bit fur­ther how we can actu­al­ly increase the accep­tance of sorts. So that was about 10% in the pre­vi­ous study. So although peo­ple found them quite accept­able, not a lot of peo­ple actu­al­ly clicked on them to accept the swaps.

Dim­itris Koutoukidis:
So what we thought was that mes­sag­ing and fram­ing may affect the choice and typ­i­cal­ly giv­ing peo­ple vague infor­ma­tion dur­ing attri­tion, qual­i­ty might not be par­tic­u­lar­ly salient. On the oth­er hand, more sell­ing strate­gies, sci­en­tists, for exam­ple, PACE or what we call the Phys­i­cal Activ­i­ty Calo­rie Equiv­a­lent, which I’ll explain in the next few slides, maybe more effec­tive. So the end of this study was to test whether incre­men­tal­ly increas­ing the tan­gi­bil­i­ty of the infor­ma­tion pro­vid­ed with low­er ener­gy swaps this time. So what we were try­ing to do was to reduce the amount of calo­ries that peo­ple order if these reduce the total ener­gy ordered in an exper­i­men­tal setting.

Dim­itris Koutoukidis:
So in this study, we ran­dom­ize peo­ple into four groups. The first one was the con­trol group where peo­ple got their stan­dard swap with no par­tic­u­lar mes­sag­ing, sim­ply say­ing, how about swap­ping your stan­dard Coke to a Coke Zero? In the sec­ond group that had the few­er calo­ries mes­sage. There was a badge say­ing that this here has few­er calo­ries. And the third group, which was the numer­ic calo­rie group were essen­tial­ly giv­ing the exact num­ber of calo­ries that peo­ple saved by mak­ing the swap. And in the final phys­i­cal activ­i­ty, calo­rie equiv­a­lent group, we were giv­ing the exact num­ber of calo­ries and how many min­utes of walk­ing this equates to.

Dim­itris Koutoukidis:
So in this par­tic­u­lar case, sav­ing 208 calo­ries, and that equates to about 46 min­utes of walk­ing. What we found was that in terms of sweet snacks, both the numer­ic calo­ries and the PACE mes­sag­ing will sig­nif­i­cant­ly reduce the ener­gy con­tent of the shop­ping bas­kets for sweet snacks. For savory snacks which were typ­i­cal­ly crisps, we saw that all inter­ven­tions were sub­stan­tial­ly reduced the ener­gy con­tent of the shop­ping bas­ket com­pared to the control.

Dim­itris Koutoukidis:
How­ev­er we did­n’t see any effect on drinks, although there was a small trend that the PACE mes­sag­ing might be a lit­tle bit more effec­tive. Now these don’t look like huge changes because if you look at the sweet snacks, that’s for 215 calo­ries on aver­age, it goes down to 200 calo­ries. But because these are only very, very small items that the length that we have to go through to offer an alter­na­tive that can be viable, the mar­gin is very small. So with­in this con­text, and if this is repli­cat­ed with mul­ti­ple snacks through­out the week, we would be able to get a clin­i­cal­ly mean­ing­ful dif­fer­ence in ener­gy intake

Dim­itris Koutoukidis:
So what are the impli­ca­tions? So of course, for research, these are exper­i­men­tal tri­als and we do need to test the strate­gies in real online set­tings, as well as study the long-term effect on food pur­chas­ing habits. But until this time comes, online super­mar­kets can def­i­nite­ly play a much more proac­tive role in shap­ing health­i­er choic­es for their cus­tomers by know­ing that both offer­ing swaps, but also alter­ing the order of prod­ucts to pri­or­i­tize the prod­ucts that are more healthy can actu­al­ly be very effective.

Dim­itris Koutoukidis:
So in sum­ma­ry, inter­ven­tions to change the food pur­chas­ing are promis­ing to nudge peo­ple into health­i­er choic­es. Alter­ing the order and alter­ing the envi­ron­ment is more effec­tive than swaps, increas­ing the tan­gi­bil­i­ty of the infor­ma­tion pro­vid­ing when offer­ing its swap increas­es the effec­tive­ness of the swab. And with­in this con­text, the phys­i­cal activ­i­ty calo­rie mes­sag­ing is the most effec­tive. I would like to thank my col­leagues and col­lab­o­ra­tors both at the Uni­ver­si­ty of Oxford and UCL and par­tic­u­lar­ly Dr. Car­men ran­dom­ized, Pro­fes­sor Susan Jebb and Sarah Breath­nach and our fun­ders and thank you for your attention.

Speak­er 2:
Amaz­ing, Dim­itris. Thank you ever so much for that. Absolute­ly fan­tas­tic. We’ve got one ques­tion in the Q and A; remem­ber if you’ve got more ques­tions, feel free to put them in. There’s one ques­tion here, basi­cal­ly, just ask­ing you to describe the con­trol group a lit­tle bit more, if you can.

Dim­itris Koutoukidis:
Sure. So essen­tial­ly in the con­trol group, what hap­pened was that we did­n’t give any inter­ven­tion at all. So par­tic­i­pants saw all prod­ucts in ran­dom order that they were to see them in the super­mar­ket. So there was actu­al­ly no influ­ence from us on the way that the prod­ucts were ranked and it was sim­ply at random.